CN106713398A

CN106713398A - Communication monitoring method and monitoring node of shared storage type cluster file system node

Info

Publication number: CN106713398A
Application number: CN201510793859.7A
Authority: CN
Inventors: 郭旭艳
Original assignee: ZTE Corp
Current assignee: ZTE Corp
Priority date: 2015-11-18
Filing date: 2015-11-18
Publication date: 2017-05-24
Also published as: WO2017084618A1

Abstract

The invention provides a communication monitoring method and monitoring node of a shared storage type cluster file system node, and relates to the field of cluster communication of shared storage type file systems. The monitoring method comprises the following steps: receiving a multicast message of a cluster node in a cluster file system, wherein the monitoring node and the cluster node are located in the cluster file system; obtaining a message type of the multicast message and a message parameter corresponding to the message type; inquiring a storage record corresponding to the multicast message in a storage table according to the message parameter; and obtaining a problem node and a cluster state of the cluster file system according to the storage record corresponding to the multicast message in the storage table when a preset time interval arrives. The scheme enables an administrator to obtain a visual observation effect on the whole cluster communication condition and to adjust the cluster communication parameters in time to improve overall performance of the cluster file system.

Description

The monitoring method and monitor node of shared memory cluster file system node communication

Technical field

The present invention relates to shared memory file system cluster communication field, more particularly to a kind of shared memory The monitoring method and monitor node of cluster file system node communication.

Background technology

As shown in figs 1 to 6, shared memory cluster file system inter-node communication module uses corosync (Corosync is a part for cluster management external member, and it can be simple by one when transmission information Configuration file come mode and agreement etc. of definition information transmission), its bottom communication is realized according to totem associations View, transmits nodal information, and being abutted against for message is ensured in the form of unicast token by way of multicast Receive, realize the synchronization of cluster interior nodes change.

When multicast message is lost, can be circulated by token and recognize and replay, until lose information node receiving Maximum is reached to the message or token cycle-indexes, Safe Order require the message broadcasted, and each is saved Will receiving for point can be just forwarded using processing, for there is Safe Order for the message of application needs broadcast Requirement message sink can unsuccessfully trigger constantly replay, cause Message Processing time delay.And Operational State is the working condition of cluster stabilization, and with stable ring, and Gather, Commit state are clusters The process of decision node state, by broadcasting itself member repeatedly, until each node member reaches consistent shape State, for the node for being confirmed to be leave, cluster also needs to isolate this trouble node, and this process can Long time can be needed, and this process cluster will not process the message of application, and such state can be led Causing the Message Processing Delays and cluster unstability of cluster increases, and existing can not in real time understand group document system System running status, it is difficult to make the problem of regulation and control to running status in time.

With the increase of cluster scale in shared memory cluster file system, time delay can be caused to increase, existed It is not adapted dynamically the parameter of clustered node to adapt to the problem the need for portfolio, it is difficult to according to actual conditions Adjustment exerts advantages of oneself.

The content of the invention

In order to overcome above mentioned problem, the present invention to provide the monitoring of shared memory cluster file system node communication Method and monitor node solve in real time understand cluster present in shared memory cluster file system File system running status, it is difficult to make the problem of regulation and control to running status in time.

In order to solve the above-mentioned technical problem, the embodiment of the present invention is adopted the following technical scheme that：

A kind of monitoring method of shared memory cluster file system node communication, is applied to shared memory collection Monitor node in group's file system, including：The multicast for receiving clustered node in the cluster file system disappears Breath, the monitor node is all located in the cluster file system with the clustered node；Obtain the multicast The type of message of message and the message parameter corresponding with the type of message；According to the message parameter, look into The stored record corresponding with the multicast message in inquiry storage table；When a prefixed time interval is reached, root According to stored record corresponding with the multicast message in the storage table, the cluster file system is obtained Trouble node and cluster state.

Alternatively, the type of message for obtaining the multicast message and corresponding with the type of message disappear Breath parameter, specifically includes：Obtain the type of message of the multicast message；When the type of message is application layer During type of message, the first message parameter of the multicast message is obtained, the first message parameter at least includes： Where message numbering, the node of multicast message described in multicast of the corresponding application layer messages of the multicast message The first transmission of first ring label of the first ring and the node of multicast message described in multicast in first ring Person address；When the type of message is that node adds type of message, obtain the multicast message second disappears Breath parameter, second message parameter at least includes：The second ring where the node of multicast message described in multicast The second ring label, second sender address of the node in second ring of multicast message described in multicast and Node member's list of the node self record of multicast message described in multicast.

Alternatively, it is described according to the message parameter, it is corresponding with the multicast message in inquiry storage table Stored record, specifically includes：When the multicast message is application layer messages type, compiled according to the message Number and the first ring label, judge in the application layer messages table of the storage table with the presence or absence of having described disappearing First stored record of breath numbering and the first ring label；It is described when not existing in the application layer messages table During the first stored record, the first message parameter of the multicast message to the application layer messages table is stored；When When there is first stored record in the application layer messages table, judged result in first ring with institute There is information drop-out in the upper node for stating the corresponding node of the first sender address, draw a upper node It is doubtful trouble node, and stores the doubtful problem in parameter to the storage table of the doubtful trouble node Node table.

Alternatively, it is described to draw a upper node for doubtful trouble node, and store the doubtful problem section Doubtful trouble node table in parameter to the storage table of point, specifically includes：According to first sender Address, obtains first node address of the doubtful trouble node in first ring；According to described first Node address, the message numbering and the first ring label, judge in the doubtful trouble node table whether In the presence of the second stored record with the first node address, the message numbering and the first ring label； When there is second stored record in the doubtful trouble node table, increase the doubtful trouble node Message Record number of times；When not existing second stored record in the doubtful trouble node table, storage bag The first parameter of the first node address, the message numbering and the first ring label is included to described doubtful Trouble node table.

Alternatively, it is described according to the message parameter, it is corresponding with the multicast message in inquiry storage table Stored record, specifically includes：When the multicast message is that node adds type of message, according to described second Ring mark and second sender address, judge whether the node in the storage table is deposited in adding message table In the 3rd stored record with the second ring label and second sender address；When the node adds When entering in message table in the absence of three stored record, the second message parameter of the multicast message is stored extremely The node adds message table；When the node is added there is three stored record in message table, root According to node member's list, judge compared with the 3rd stored record, the node of the multicast message into With the presence or absence of the node member for increasing or decreasing in member's list；When in node member's list of the multicast message During in the presence of the node member for reducing, the Section Point address of the node member of the reduction is obtained, according to described Section Point address and second sender address, judge that whether being left for the storage table deposit in node table In the 4th stored record with the Section Point address and second sender address；Left when described When there is four stored record in node table, increase the Message Record number of times of the node member of the reduction； When it is described leave node table in the absence of four storage information when, storage include the Section Point address, Second parameter of second sender address leaves node table to described；When the multicast message node into When there is increased node member in member's list, the 3rd node address of the increased node member is obtained, According to the 3rd node address and second sender address, judge described in leave and whether deposit in node table In the 5th stored record with the 3rd node address and second sender address；Left when described When there is five stored record in node table, the 5th stored record is deleted.

Alternatively, it is described when a prefixed time interval is reached, disappear with the multicast according in the storage table The corresponding stored record of manner of breathing, obtains the trouble node and cluster state of the cluster file system, specific bag Include：In the doubtful trouble node table, reached when there are Message Record number of times in the doubtful trouble node During the especially doubtful trouble node of preset maximum value, judge that the preset maximum value especially doubtful is asked with except described Whether the Message Record number of times of other nodes outside topic node is identical；When preset maximum value is especially doubted with except described Like other nodes outside trouble node Message Record number of times it is identical when, judge the cluster file system be collection Group business busy state；When disappearing for preset maximum value and other nodes in addition to the especially doubtful trouble node When breath record number of times is different, judge that the especially doubtful trouble node is described problem node.

Alternatively, it is described when a prefixed time interval is reached, disappear with the multicast according in the storage table The corresponding stored record of manner of breathing, obtains the trouble node and cluster state of the cluster file system, specific bag Include：Leave whether node table is empty described in judging；To leave node table be not sky when described, and described leaves section In point table, exist it is multiple have described in identical during the stored record of Section Point address, judge and the phase The corresponding node in the same Section Point address is described problem node；When the node table that leaves is sky, In judging that the node adds message table, with second sender address minimum in second ring Whether the quantity of stored record reaches preset value；In the node adds message table, with second ring When the quantity of the stored record of middle minimum second sender address reaches preset value, the cluster is judged File system is token time-out frequency.

Alternatively, the monitoring method also includes：Obtain corresponding according to described problem node and cluster state Adjusting parameter；By the adjusting parameter multicast to the clustered node, so that the clustered node is according to Current itself configuration of adjusting parameter adjustment.

Alternatively, it is described that corresponding adjusting parameter, specific bag are obtained according to described problem node and cluster state Include：It is pre- according to first when the cluster file system residing for the clustered node is group service busy state If multiplying power tune up current message transmission window value for a new information transmission window value, according to second default times Rate reduce each clustered node current maximum can transmission information value be that a new maximum can transmission information Value；It is pre- according to the 3rd when the cluster file system residing for the clustered node is token time-out frequency If multiplying power to tune up the token time-out time of the clustered node be a new token time-out time.

A kind of monitor node in shared memory cluster file system, including：

First receiver module, the multicast message for receiving clustered node in the cluster file system is described Monitor node is all located in the cluster file system with the clustered node；First acquisition module, for obtaining Take the type of message and the message parameter corresponding with the type of message of the multicast message；Enquiry module, For according to the message parameter, inquiring about stored record corresponding with the multicast message in storage table；The Two acquisition modules, for when a prefixed time interval is reached, disappearing with the multicast according in the storage table The corresponding stored record of manner of breathing, obtains the trouble node and cluster state of the cluster file system.

Alternatively, first acquisition module specifically for：Obtain the type of message of the multicast message；When When the type of message is application layer messages type, the first message parameter of the multicast message is obtained, it is described First message parameter at least includes：Message numbering, the multicast institute of the corresponding application layer messages of the multicast message First ring label of the first ring where stating the node of the multicast message and node of multicast message exists described in multicast The first sender address in first ring；When the type of message is that node adds type of message, obtain The second message parameter of the multicast message is taken, second message parameter at least includes：Multicast described in multicast Second ring label, the node of multicast message described in multicast of the second ring where the node of message are described second Node member's list of the node self record of the second sender address in ring and multicast message described in multicast.

Alternatively, the enquiry module is specifically included：

First judging submodule, for when the multicast message is application layer messages type, being disappeared according to described Breath numbering and the first ring label, judge to whether there is with institute in the application layer messages table of the storage table State the first stored record of message numbering and the first ring label；First sub-module stored, for when described When not existing first stored record in application layer messages table, the first message ginseng of the multicast message is stored Count to the application layer messages table；Second sub-module stored, for when there is institute in the application layer messages table When stating the first stored record, judged result is corresponding with first sender address in first ring There is information drop-out in a upper node of node, draw a upper node for doubtful trouble node, and store institute State the doubtful trouble node table in parameter to the storage table of doubtful trouble node.

Alternatively, second sub-module stored is specifically included：Acquiring unit, for according to the described first hair The person of sending address, obtains first node address of the doubtful trouble node in first ring；Judging unit, For according to the first node address, the message numbering and the first ring label, judging described doubtful With the presence or absence of with the first node address, the message numbering and the first ring mark in trouble node table Number the second stored record；Recording unit, for depositing when having described second in the doubtful trouble node table During storage record, increase the Message Record number of times of the doubtful trouble node；Memory cell, for being doubted when described During like not existing second stored record in trouble node table, storage includes the first node address, institute State the first parameter of message numbering and the first ring label to the doubtful trouble node table.

Alternatively, the enquiry module is specifically included：

Second judging submodule, for when the multicast message is that node adds type of message, according to described Second ring mark and second sender address, judge that the node in the storage table is in adding message table No the 3rd stored record existed with the second ring label and second sender address；3rd storage Submodule, for when the node is added and do not exist three stored record in message table, storage to be described Second message parameter of multicast message to the node adds message table；3rd judging submodule, for working as State node to add in message table when there is three stored record, according to node member's list, judge Compared with the 3rd stored record, added deduct with the presence or absence of increasing in node member's list of the multicast message Few node member；4th judging submodule, for existing when in node member's list of the multicast message During the node member of reduction, the Section Point address of the node member of the reduction is obtained, according to described second Node address and second sender address, judge leaving in node table with the presence or absence of tool for the storage table There is the 4th stored record of the Section Point address and second sender address；Record sub module, uses In when it is described leave node table in there is four stored record when, increase the node member's of the reduction Message Record number of times；4th sub-module stored, for when it is described leave node table in deposited in the absence of the described 4th During storage information, storage includes the second parameter of the Section Point address, second sender address to institute State and leave node table；5th judging submodule, for existing when in node member's list of the multicast message During increased node member, the 3rd node address of the increased node member is obtained, according to the described 3rd Node address and second sender address, judge that described being left whether there is in node table with described the 5th stored record of three node address and second sender address；Submodule is deleted, for when described Leave in node table when there is five stored record, delete the 5th stored record.

Alternatively, second acquisition module, specifically includes：

6th judging submodule, for the doubtful trouble node table in, deposited when in the doubtful trouble node When there is Message Record number of times to reach the especially doubtful trouble node of preset maximum value, the default maximum is judged Whether value is identical with the Message Record number of times of other nodes in addition to the especially doubtful trouble node, when default When maximum is identical with the Message Record number of times of other nodes in addition to the especially doubtful trouble node, judge The cluster file system is group service busy state；7th judging submodule, for working as preset maximum value When Message Record number of times from other nodes in addition to the especially doubtful trouble node is different, the spy is judged Not doubtful trouble node is described problem node.

Alternatively, second acquisition module, specifically includes：

8th judging submodule, for judging described to leave whether node table is empty；9th judging submodule, For leaving node table and not being sky when described, and it is described leave node table, existing multiple has identical institute When stating the stored record of Section Point address, judge corresponding with Section Point address described in the identical Node is described problem node；Tenth judging submodule, for being sky when the node table that leaves, judges institute State in node addition message table, the storage with second sender address minimum in second ring is remembered Whether the quantity of record reaches preset value；11st judging submodule, in adding message table when the node, The quantity of the stored record with second sender address minimum in second ring reaches preset value When, judge that the cluster file system is token time-out frequency.

Alternatively, the monitor node also includes：

3rd acquisition module, for obtaining corresponding adjusting parameter according to described problem node and cluster state； Multicast module, for by the adjusting parameter multicast to the clustered node so that the clustered node according to Current itself configuration of adjusting parameter adjustment.

Alternatively, the 3rd acquisition module, specifically for：

It is pre- according to first when the cluster file system residing for the clustered node is group service busy state If multiplying power tune up current message transmission window value for a new information transmission window value, according to second default times Rate reduce each clustered node current maximum can transmission information value be that a new maximum can transmission information Value；It is pre- according to the 3rd when the cluster file system residing for the clustered node is token time-out frequency If multiplying power to tune up the token time-out time of the clustered node be a new token time-out time.

The beneficial effects of the invention are as follows：

Such scheme, cluster operation conditions is monitored by collecting multicast message, and according to the group of each node Message statistics analysis node state is broadcast, Statistic analysis are given to cluster state and trouble node, improve its communication Traffic handing capacity and stability, and equipment fault notice can be obtained the very first time, make administrative staff to whole Individual cluster communication situation has observing effect directly perceived, can understand in time status of equipment, positioning fault target, Operating efficiency is improved, and then improves cluster file system overall performance.

Brief description of the drawings

Fig. 1 is the application A1 message M1M2M3 that the corosync on node P1 receives on the node, The message schematic diagram of multicast in cluster；

Fig. 2 be node multicast message after, the schematic diagram that token token is passed in P2 by node P1；

After Fig. 3 receives token confirmation reception message for P2, continue to transmit the schematic diagram of token to P3；

Fig. 4 is node broadcasts node addition message joinmsg schematic diagrames after addition node in cluster；

Fig. 5 is to broadcast itself member set's schematic diagram after cluster other nodes receive joinmsg；

Fig. 6 is that node does not receive other nodes joinmsg so as to the schematic diagram of unrealized consensus；

Fig. 7 is increase monitor node schematic diagram in cluster；

Fig. 8 is method flow diagram in first embodiment of the invention；

Fig. 9 is method flow diagram in second embodiment of the invention；

Figure 10 is method flow diagram in third embodiment of the invention；

Figure 11 is method flow diagram in fourth embodiment of the invention；

Figure 12 is method flow diagram in fifth embodiment of the invention；

Figure 13 is method flow diagram one in sixth embodiment of the invention；

Figure 14 is method flow diagram two in sixth embodiment of the invention；

Figure 15 is method flow diagram one in seventh embodiment of the invention；

Figure 16 is method flow diagram two in seventh embodiment of the invention；

Figure 17 is overall structure block diagram in ninth embodiment of the invention；

Figure 18 is detailed block diagram in ninth embodiment of the invention；

Figure 19 is the overall flow schematic diagram one of method in the present invention；

Figure 20 is the overall flow schematic diagram two of method in the present invention.

Specific embodiment

The exemplary embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although being shown in accompanying drawing The exemplary embodiment of the disclosure, it being understood, however, that may be realized in various forms the disclosure without should be by Embodiments set forth here is limited.Conversely, there is provided these embodiments are able to be best understood from this It is open, and can by the scope of the present disclosure it is complete convey to those skilled in the art.

First embodiment

As shown in Figure 7, Figure 8, the present invention provides a kind of shared memory cluster file system node communication Monitoring method, is applied to the monitor node in shared memory cluster file system, and the method includes：

Step 101：Receive the multicast message of clustered node in the cluster file system.

Here can be by the cluster multicast group of monitor node addition cluster file system so that the monitor node It is all located in cluster file system with above-mentioned clustered node, wherein the monitor node can be and clustered node phase Same main frame or blade server or other servers.

Step 102：Obtain the type of message and the message corresponding with the type of message of the multicast message Parameter.

Receive in a step 101 after multicast message, just obtain the corresponding type of message of the multicast message, And then different message parameter corresponding from different type of messages is obtained according to the type of message.

Step 103：It is corresponding with the multicast message in inquiry storage table to deposit according to the message parameter Storage record.

According to the message parameter obtained in step 102, the pass corresponding with message parameter is inquired about from storage table In multicast message associated storage record, with step 101 accordingly, the storage table is every to each node of cluster Multicast message once can all be collected and count.

Step 104：When a prefixed time interval is reached, according in the storage table with the multicast message Corresponding stored record, obtains the trouble node and cluster state of the cluster file system.

Storage table can a period of time in the multicast message in cluster file system is acquired with record, When presetting time interval is reached, judged according to the stored record in storage table within the time period and obtained Whether the cluster state for taking cluster file system is what kind of and has the trouble node of correlation to occur.

The monitoring method is to increase by a node in existing cluster, node configuration cluster multicast address, can Receive cluster broadcast message.By receiving the multicast message of each node in cluster file system, disappeared according to multicast The type of breath and the message parameter corresponding with multicast message, on being deposited to multicast message in inquiry storage table Storage records to know the trouble node and cluster state of cluster file system, in shared memory group document Cast communication message analysis node running status is utilized in system, is overcome in the prior art and be there is shared storage In the peer-to-peer framework of formula cluster file system, cluster disposal ability, node failure lack statistical analysis and collect And the problem and defect that collection swarm parameter can not be adjusted dynamically.

Second embodiment

Specifically, as shown in figure 9, on the basis of first embodiment, the acquisition multicast message in step 102 Type of message and the message parameter corresponding with the type of message, specifically include：

Step 1021：Obtain the type of message of the multicast message.

Step 1022：When the type of message is application layer messages type, the multicast message is obtained First message parameter.

The first message parameter at least includes：Message numbering, the group of the corresponding application layer messages of the multicast message The node of the first ring label of the first ring where broadcasting the node of the multicast message and the multicast multicast message exists The first sender address in first ring.

Step 1023：When the type of message is that node adds type of message, the multicast message is obtained The second message parameter.

Second message parameter at least includes：Second ring of the second ring where the node of the multicast multicast message Label, second sender address of the node in second ring of the multicast multicast message and the multicast multicast disappear Node member's list of the node self record of breath.

Multicast message is received, the type of message of the multicast message is obtained, type of message here is broadly divided into Two classes a, class is application layer messages type, and a class is that node adds message joinmsg types, wherein applying Layer type of message refers to the application layer messages for sending application layer when having the message that application layer is sent in multicast ring Type of message in multicast to cluster during other each nodes, it refers in multicast ring that its interior joint adds type of message In when thering is node to add, between each node multicast added on the node and self record node member's collection The type of message of conjunction.When the multicast message is application layer messages type, the parameter of acquisition will at least include： Where the message numbering seq of the corresponding application layer messages of the multicast message, the node of the multicast multicast message The node of the first ring label ring_id1 of the ring label of ring, i.e. the first ring and the multicast multicast message is in ring Address, i.e. the first sender address sender_id1 in the first ring；When type of message is that node addition disappears During breath type, the parameter of acquisition at least includes：The ring label of the ring where the node of the multicast multicast message, That is the address of the node in ring of the second ring label ring_id2 of the second ring, the multicast multicast message, i.e., The section of the node self record of the second sender address sender_id2 in two rings and the multicast multicast message The member set of point member, i.e. node member's list proc_list.To the acquisition of above parameter in order to group The cluster state that message gives expression to is broadcast to judge.

3rd embodiment

Further, as shown in Figure 10, on the basis of first embodiment and second embodiment, multicast is worked as When message is application layer messages type, according to the message parameter in step 103, with group in inquiry storage table The corresponding stored record of message is broadcast, is specifically included：

Step 1031：When multicast message is application layer messages type, according to the message numbering and described First ring label, judge in the application layer messages table of the storage table with the presence or absence of have the message numbering and First stored record of the first ring label.

Here, the relevant parameter being stored with the storage table to the multicast message of clustered node and the multicast message Record, wherein storage table include application layer messages table, the application layer messages table correspondence storage be on group Relative recording when broadcasting message for application layer messages type.When multicast message is application layer messages type, root According to seq and ring_id1 corresponding with the multicast message, judge whether existed in the application layer messages table It is that multicast in same ring to same application-level message is recorded to have with the multicast message, i.e., the first above-mentioned storage Record.

Step 1032：When not existing first stored record in the application layer messages table, institute is stored State the first message parameter of multicast message to the application layer messages table.

According to the judgement in step 1031, in the absence of the first storage in judged result is application layer messages table During record, the multicast message is just stored in application layer messages table, storage is up to major general and the multicast message pair The first message parameter answered is stored in application layer messages table.

Step 1033：When there is first stored record in the application layer messages table, judged result Lost for a upper node of node corresponding with first sender address in first ring has message Lose, show that a upper node is doubtful trouble node, and store the parameter of the doubtful trouble node to institute State the doubtful trouble node table in storage table.

Also include doubtful trouble node table in the storage table, according to the judgement in step 1031, when judgement is tied Fruit is when there is the first stored record in application layer messages table, to show on same ring to identical application layer messages The situation of repetition multicast is occurred in that, can thus be drawn and triggered this heavy with the presence of node information drop-out on the ring The situation of multiple multicast, with reference in shared memory cluster file system node communication routine, token in multicast ring The message confirming process of token, judges that it is Pn to retransmit information node, and the order according to token alternative spaces is pushed away The Pn-1 nodes surveyed in current member list are classified as doubtful trouble node to lose information node, are obtained Know that a upper node of the node corresponding with the sender_id1 in the multicast message has information drop-out, and then Judge to be likely to occur problem on this at a node, as doubtful trouble node, by the doubtful trouble node and Relevant parameter is stored to doubtful trouble node table.

Fourth embodiment

Further, as shown in figure 11, a section is drawn on the basis of 3rd embodiment, in step 1033 Point is doubtful trouble node, and doubtful asking of storing in parameter to the storage table of the doubtful trouble node Topic node table, specifically includes：

Step 10331：According to the first sender address, the doubtful trouble node is obtained in first ring In first node address.

In with reference to shared memory cluster file system node communication routine, token tokens disappears in multicast ring Breath confirmation process, learns that a upper node of the node corresponding with the sender_id1 in the multicast message is present Information drop-out, and then judge to be likely to occur problem on this at a node, and first sender address is group Broadcast address of the node of the multicast message in the first ring, the node address in multicast ring be by sequence number from it is small to Big sequential organization, can learn that a node exists on the node according to first sender address sender_id1 Address in first ring, i.e., above-mentioned first node address nodeid1.

Step 10332：According to the first node address, the message numbering and the first ring label, Judge in the doubtful trouble node table with the presence or absence of have the first node address, the message numbering and Second stored record of the first ring label.

Nodeid1 is obtained in step 10331, according to nodeid1, seq and the ring_id1, is judged With the presence or absence of same node loss same application-level in same ring in the record stored in doubtful trouble node table The record of message, i.e. second stored record.

Step 10333：When there is second stored record in the doubtful trouble node table, increase institute State the Message Record number of times of doubtful trouble node.

It is existing in the judged result in step 10332 is doubtful trouble node table to have the second stored record When, now increase the record number of times for losing same application-level message in same ring to the doubtful trouble node, Namely increase in same ring by the doubtful trouble node exist message loss and other node weights for causing The number of times of multiple multicast.

Step 10334：When not existing second stored record in the doubtful trouble node table, storage The first parameter including the first node address, the message numbering and the first ring label is doubted to described Like trouble node table.

Not existing in the judged result in step 10332 is doubtful trouble node table has the second stored record When, at least the parameters such as nodeid1, seq and ring_id1 are stored to doubtful trouble node table, which is recorded In individual ring which node exist lose which application layer messages the occurrence of.

5th embodiment

As shown in figure 12, on the basis of first embodiment and second embodiment, when multicast message is node When adding type of message, according to the message parameter in step 103, disappear with the multicast in inquiry storage table The corresponding stored record of manner of breathing, specifically includes：

Step 1034：When the multicast message is that node adds type of message, according to the second ring mark Note and second sender address, judge that the node in the storage table whether there is tool in adding message table There is the 3rd stored record of the second ring label and second sender address.

Also include that node adds message table in storage table, when the multicast message is that node adds message, according to Ring_id2, sender_id2, judge the node add message table in whether be stored with the multicast message be The node that same node sends in same ring adds the Message Record of type of message, i.e., the 3rd above-mentioned storage note Record.

Step 1035：When the node is added does not exist three stored record in message table, storage Second message parameter of the multicast message to the node adds message table.

The node of same node multicast adds message class in not existing to same ring during node adds message table When the record of the message of type is three stored record, just the multicast message is stored to node and adds message table, Storage is stored in application layer messages table up to major general's the second message parameter corresponding with the multicast message.

Step 1036：When the node is added there is three stored record in message table, according to institute Node member's list is stated, is judged compared with the 3rd stored record, node member's row of the multicast message With the presence or absence of the node member for increasing or decreasing in table.

When node is added there is three stored records in message table, show that the node is stored in adding message table There is the Message Record that type of message is added with the node that the multicast message is the same node transmission in same ring, Deposited according in the proc_list in the second message parameter corresponding with multicast message, with the 3rd stored record Node member's list parameter compare, judge in the proc_list of the multicast message with the presence or absence of increasing or The node member of reduction, that is, judge the multicast message multicast on the phase leaving or add when front ring interior joint Concern feeling condition.

Step 1037：When there is the node member for reducing in node member's list of the multicast message, The Section Point address of the node member of the reduction is obtained, according to the Section Point address and described second Sender address, judges leaving in node table with the presence or absence of having the Section Point address for the storage table And the 4th stored record of second sender address.

Also include leaving node table (leave tables) in storage table, when the judged result of step 1036 is multicast There is the node member for reducing in node member's list of message, i.e., there is the node member with the reduction in ring The corresponding node for leaving, then obtain Section Point address of the node member of the reduction in the second ring Whether nodeid2, judges that this leaves and be stored with node table same node according to nodeid2, sender_id2 The Message Record of the node with same reduction for sending, i.e., above-mentioned 4th stored record.

Step 1038：When it is described leave node table in there is four stored record when, increase described in subtract The Message Record number of times of few node member.

When judged result in leaving node table to have four stored records in step 1037, show that this subtracts Few node member now increases corresponding by same node to the excessively corresponding message of other clustered node multicasts Message Record number of times, that is, increase the record for leaving number of times to the node member of the reduction.

Step 1039：When it is described leave node table in the absence of four storage information when, storage includes The Section Point address, the second parameter of second sender address leave node table to described.

When judged result is in leaving node table in the absence of four stored records in step 1037, just to phase Parameter is answered to be stored, storage is stored in up to parameters such as major general's Section Point address, the second sender address This leaves node table.

Step 10310：When there is increased node member in node member's list of the multicast message, The 3rd node address of the increased node member is obtained, according to the 3rd node address and described second Sender address, judges that described being left whether there is in node table with the 3rd node address and described the 5th stored record of two sender address.

When step 1036 judged result for multicast message node member's list in exist increased node into Member, i.e., the node that there is addition corresponding with the increased node member in ring, then obtain the increased node Threeth node address nodeid3 of the member in the second ring, being judged according to nodeid3, sender_id2 should be from Open whether be stored with node table on this increased node member by same node send to the section The record of the message left before point, i.e., above-mentioned 5th stored record.

Step 10311：When it is described leave node table in there is five stored record when, delete described the Five stored records.

Exist in node table is left sent by same sending node with this increased node member this into The record left before member, then delete the record.

In above-mentioned step 1034 to step 10311, to when multicast message be node add type of message when, It is corresponding with multicast message in inquiry storage table according to the message parameter corresponding with the type of message is obtained The condition judgment for using and carrying out of the relevant parameter that the process of stored record and centre are related to is made that description, During message table is added to the node in storage table, node table is left by using the parameter of the multicast message for obtaining Data with existing make multilevel iudge, corresponding parameter stored according to correlated results finally or added up or The corresponding operating of deletion, reaches the real-time collecting to cluster file system status information and monitoring.

Sixth embodiment

The different messages type of multicast message will be directed in the present embodiment, to how to obtain cluster file system Trouble node and cluster state make description.

As shown in figure 13, on the one hand, when the type of message of multicast message is application layer messages type, base In first embodiment and 3rd embodiment, in step 104 when a prefixed time interval is reached, according to storage The stored record corresponding with multicast message in table, obtains the trouble node and cluster state of cluster file system, Specifically include：

Step 1041：In the doubtful trouble node table, when there are message in the doubtful trouble node When record number of times reaches the especially doubtful trouble node of preset maximum value, judge the preset maximum value and remove institute Whether the Message Record number of times for stating other nodes outside especially doubtful trouble node is identical.

Step 1042：When disappearing for preset maximum value and other nodes in addition to the especially doubtful trouble node When breath record number of times is identical, judge that the cluster file system is group service busy state.

Step 1043：When disappearing for preset maximum value and other nodes in addition to the especially doubtful trouble node When breath record number of times is different, judge that the especially doubtful trouble node is described problem node.

When the type of message of multicast message is application layer messages type, and predetermined time interval is reached, and is being sentenced In disconnected cluster file system whether problematic node and during current cluster state, it is necessary to doubtful trouble node Data in table are judged and are analyzed, disappeared when having in the doubtful trouble node stored in doubtful trouble node table The cumulative number for ceasing record reaches the node appearance of preset maximum value, that is, when there is especially doubtful trouble node, Need to judge especially doubtful trouble node Message Record number of times interior at preset time intervals reaches this most Big value whether with the Message Record number of times of other doubtful trouble nodes that records in doubtful trouble node table whether Identical, if identical, that just can consider that the doubtful trouble node in clustered node now is all occurred in that together Information drop-out more than sample, then judge that cluster file system is in group service busy state, if it is not the same, That just can consider that the especially doubtful trouble node for reaching Message Record number of times maximum is present and frequently disappears Breath loss situation, you can judge that the node is trouble node.

As shown in figure 14, on the other hand, when the type of message of the multicast message is node addition type of message When, based on first embodiment and the 5th embodiment, in step 104 when a prefixed time interval is reached, root According to stored record corresponding with multicast message in storage table, the trouble node and collection of cluster file system are obtained Group's state, specifically includes：

Step 1044：Leave whether node table is empty described in judging.

Step 1045：When described leave node table and be not sky, and it is described leave node table, there is multiple With during the stored record of Section Point address, judging described in identical and Section Point described in the identical The corresponding node in address is described problem node.

When the type of message of multicast message is node addition type of message, the problem in cluster file system is obtained When node and cluster state, first to be judged and analyzed to leaving the data collected in node table and store, With the presence of record in node table is left, i.e., it is not space-time, and in the stored record for leaving node table, deposit When this parameter of identical nodeid2 is all included in having multiple record, then different storage notes are would know that It is corresponding in record to have the record that node is reduced to identical, the node corresponding with the nodeid2 can be learnt Occur repeatedly exiting the situation generation for leaving multicast ring, then now judge the section corresponding with the nodeid2 Point is trouble node.

Step 1046：When the node table that leaves is sky, judge in the node addition message table, have Whether the quantity of the stored record of minimum second sender address reaches preset value in second ring.

Step 1047：In the node adds message table, with second ring minimum described the When the quantity of the stored record of two sender address reaches preset value, judge that the cluster file system is token Overtime frequency.

Here, the second minimum sender address refers in multicast ring in order in the node address of layout, The address of node for sending multicast message to other nodes is minimum in the node address of the ring.When leaving section Exist without record in point table, as space-time is, it is necessary to the node in judging storage table adds the storage of message table In record, the quantity of the stored record of the second sender address parameter with the minimum is default either with or without reaching Value, i.e., by the node multicast corresponding with the second sender address of the minimum multicast message quantity whether Preset value is reached, the preset value is generally the default maximum in allowed band, when preset value is reached, Illustrate the node corresponding with the second sender address of the minimum frequent transmission group within a preset time interval Broadcast message, cause this it is frequent send operation the reason for be generally transmission multicast message to other nodes after, Token tokens are responded without timely being received by other nodes, therefore are caused and frequently retransmitted, then herein When judge the cluster file system be in token time-out frequency.

The above-mentioned type according to multicast message is carried out in different inquiry and deterministic processes in storage table, is passed through Comparing between the stored counts of longitudinal node in itself and horizontal node, judges cluster state and trouble node, and When the effective state to cluster file system make detection and judge.

7th embodiment

Describe in the first embodiment how real-time monitoring and obtain cluster file system trouble node and Cluster state, as shown in figure 15, to monitoring and will obtain the problem of cluster file system in the present embodiment After node and cluster state, how to solve the problems, such as to be described in cluster file system, correspondingly, should Monitoring method also includes：

Step 105：Corresponding adjusting parameter is obtained according to described problem node and cluster state.

Step 106：By the adjusting parameter multicast to the clustered node so that the clustered node according to Current itself configuration of adjusting parameter adjustment.

When step 104 obtain cluster file system trouble node and cluster state after, according to the problem Node and cluster state provide corresponding countermeasure, now by obtaining targetedly adjusting parameter, And by the adjusting parameter multicast to clustered node, other nodes is adjusted in time according to the adjusting parameter and itself match somebody with somebody Put to solve system problem.

Further, as shown in figure 16, obtained according to trouble node and cluster state in above-mentioned steps 105 Corresponding adjusting parameter, specifically includes：

Step 1051：When the cluster file system residing for the clustered node is group service busy state, Current message transmission window value is tuned up for a new information transmission window value according to the first default multiplying power, according to the Two default multiplying powers reduce each clustered node current maximum can transmission information value be a new maximum Can transmission information value.

Here the first default multiplying power is preferably 1.2 multiplying power, and the second default multiplying power is preferably 0.9 Multiplying power.According to the particular problem point situation treatment that cluster occurs, in the state of group service is busy, each Node has multicast message to lose, and now tunes up message transmission window value window_size according to 1.2 multiplying power, Maximum when each clustered node multicast message is reduced according to 0.9 multiplying power can transmission information value Max_messages, to realize alleviating cluster busy state, reduces information drop-out situation.

Step 1052：When the cluster file system residing for the clustered node is token time-out frequency, When the token time-out time for tuning up the clustered node according to the 3rd default multiplying power is a new token time-out Between.

Here the 3rd default multiplying power is preferably 1.2 multiplying power, for the frequent problem of token tokens time-out, The token time-out time of each node in cluster is tuned up according to 1.2 multiplying power, to reduce in group system because of order The message retransmission that board time-out is caused.

Correspondingly, after the trouble node and cluster state of acquisition cluster file system, can be to including occurring asking Topic node, token tokens time-out frequently, group service it is busy interior every kind of result and with various result phases Corresponding adjusting parameter carries out output prompting.

Further, the corosync in cluster management external member is distributed in as the core component of cluster communication On cluster each node, the parameter configuration of corosync is identical in each clustered node, and it can be by this node Application message be broadcast in cluster, and receive the multicast message of other nodes, whole cluster shape is reached with this The synchronization of state, and the transmission for passing through token, realize the reliable transmission of message, and cluster member change is detected, from And maintain whole cluster stabilization.In the method mentioned in the present invention, can in the cluster in addition to monitor node its The communication layers of his node corosync increase regulation parameter message, after corosync receives the message, carry out The totem configuration modifications of corosync.By the corosync in the cluster management external member in other clustered nodes Receive the parsing of parameter adjustment message and relevant configuration is modified.

8th embodiment

To make the object of the invention, technical scheme advantage clearer, the present embodiment will be further combined with specific reality Live scape is illustrated.

The first step：Monitor node adds multicast group, prepares receiving multicast message.

Second step：Receive multicast message, parse multicast message, at least obtain type of message and with message class The corresponding supplemental characteristic of type, and those data are preserved.

3rd step：According to the Query Result of supplemental characteristic, corresponding data, decision node state are stored.

4th step：Timing carries out statistical analysis, inquires about trouble node table and leaves node table, judges possible Trouble node, cluster state, according to adjusting parameter is currently configured, output statistics and investigation are advised.

5th step：Adjusting parameter is broadcasted in the cluster.

6th step：Clustered node receives the message on adjusting parameter, is updated according to adjusting parameter and configured.

Using method disclosed by the invention, compared with prior art, cluster and the treatment of each cluster partial node are realized The intuitively observation effect of ability, configuration parameter is dynamically adjusted according to cluster service requirement, improves cluster Reliability and stability, improve the observation method of cluster communication, facilitate developer to analyze.

Wherein, storage table specifically includes following sections：

Application layer messages table, can be comprising row sender_id1, seq, srpaddr, ring_id1, timestamp.

Doubtful trouble node table, can be comprising nodeid1, seq, ring_id1, timestamp.

Node adds message table, can be comprising row sender_id2, ring_id2, proc_list, timestamp.

Leave tables, can be comprising sender_id2, timestamp and not present in proc_list nodeid2。

Wherein srpaddr arranges the physical address for recording multicast message sending node, and timestamp is arranged to be used for The correlation time of multicast message is recorded, for the record to the time, in order to carry out in a prefixed time interval During arrival, the stored record in storage table obtains the trouble node and cluster state of cluster file system, Can also realize, when a preset time period is reached, periodically removing the data in storage table by timing.

Further, the monitoring that will be communicated to shared memory cluster file system node with reference to Figure 19 here In method, the multicast message of clustered node in cluster file system is received, obtain the type of message of multicast message And the message parameter corresponding with type of message, according to the message parameter, disappear with the multicast in inquiry storage table The process of the corresponding stored record of manner of breathing makes whole description.

After process starts, receive the multicast message of clustered node in cluster file system to judge the multicast Whether the type of message of message is application layer messages type, if so, then parsing obtain the multicast message seq, Sender_id1, ring_id1, the application layer messages of storage table are inquired about with seq and ring_id1 as querying condition Whether there is relative recording in table, if not existing relative recording in application layer messages table, just store the message Into application layer messages table, if there is relative recording in application layer messages table, just according in the relative recording Sender of the message address, judges a upper sections of the node Pn corresponding with the sender of the message address in ring Point Pn-1 is doubtful trouble node, now judges to whether there is with node in doubtful trouble node table again The node address nodeid and the relative recording with identical seq, ring_id1 of Pn-1, if in the presence of, The record number of times of the relative recording of the node address nodeid with Pn-1 is updated, that is, updates the nodeid What the information drop-out of corresponding Pn-1 nodes was caused does not receive message number of times, if not existing, preserves Pn-1's Doubtful trouble node table in nodeid, seq, ring_id1 to storage table；If the multicast message is not application During layer type of message, judge whether the multicast message is joinmsg type of messages, if so, then obtaining the group Sender_id2, ring_id2, proc_list parameter of message are broadcast, with sender_id2 and ring_id2 as bar Node in part inquiry storage table adds whether message table is stored with identical recordings, disappears if then comparing multicast Proc_list in breath and the identical recordings, when the node being reduced in the proc_list of the multicast message, obtains The node address nodeid of the node of the reduction is taken, now again with parameter sender_id2, the node of the reduction Node address nodeid, ring_id2 inquiry leave tables in whether there is relative recording, if leave tables In there is relative recording, then the record number of times of the nodeid corresponding nodes of adding up, that is, add up the nodeid Leave number of corresponding node, if not existing relative recording in leave tables, increases in leave tables To the record of the message of the node of the corresponding reductions of the nodeid；When multicast message is neither application layer messages Type is also not joinmsg type of messages, then abandon the multicast message.

Correspondingly, the monitoring side that will be communicated to shared memory cluster file system node with reference to Figure 20 here In method, according to the message parameter corresponding with the type of message of multicast message, disappear with multicast in inquiry storage table The corresponding stored record of manner of breathing；When a prefixed time interval is reached, according in storage table with multicast message phase Corresponding stored record, the process of the trouble node and cluster state that obtain the cluster file system makes entirety Description.

When prefixed time interval is reached, doubtful trouble node table is checked, count each doubtful trouble node The number of times of the appearance mistake of record number of times, i.e. each doubtful trouble node, judge in doubtful trouble node whether The Message Record number of times for having node has reached default maximum, if it has, the Message Record of the node is secondary Number is made comparisons with the Message Record number of times of other doubtful trouble nodes, when the message of other doubtful trouble nodes is remembered Record number of times is identical with the record number of times of the node that Message Record number of times reaches default maximum, then judge cluster Heavy traffic, if differing, judges that Message Record number of times reaches the node of default maximum and there is problem； When preset maximum value is reached in the absence of the Message Record number of times for having node in doubtful trouble node, then inquire about Whether leave tables, judge with the presence of record in leave tables, if with the presence of node in leave tables, judging this With the presence or absence of multiple stored records with identical node address in the record of leave tables, i.e., in leave tables Record whether be the Message Record left to same node, if so, then showing and the identical node address There is problem, it is necessary to wait cluster ruling to process in corresponding node；When no record is present in leave tables, Then whether decision node reaches in adding message table on the Message Record with the node of lowest address in ring Preset value, if so, there is token time-out in then judging cluster frequently, if it is not, not entering to the situation then Row treatment.

The present invention monitors cluster and node state by receiving and analyzing each node multicast message of cluster, and right Cluster is busy, and token timeout cases are adjusted to adapt to cluster treatment by multicast to each node parameter Ability, corosync communication modules are not set on the monitor node for adding multicast group, to carry out parameter regulation, are needed To increase by a regulation parameter Message Processing in clustered node communication module, but not influence existing cluster scale, The monitor node is not involved in specific business, monitors cluster operation conditions by collecting multicast message, and according to The multicast message statistical analysis node state of each node, Statistic analysis are given to cluster state and trouble node, And adjustment collects swarm parameter in good time, its communication service processing ability and stability is improved, to whole cluster communication shape Condition has observing effect directly perceived, can obtain equipment fault notice the very first time, enables administrative staff timely Understand status of equipment, positioning fault target, improve operating efficiency.

9th embodiment

As shown in Fig. 7, Figure 17, the invention also discloses in a kind of shared memory cluster file system Monitor node, the monitor node includes：First receiver module 2100, the first acquisition module 2200, inquiry Module 2300, the second acquisition module 2400.

Wherein, the first receiver module 2100, the multicast for receiving clustered node in cluster file system disappears Breath, the monitor node is all located in cluster file system with clustered node；First acquisition module 2200, uses In the type of message and the message parameter corresponding with the type of message that obtain the multicast message；Enquiry module 2300, for according to the message parameter, inquiring about stored record corresponding with the multicast message in storage table； Second acquisition module 2400, for when a prefixed time interval is reached, according in the storage table with the group The corresponding stored record of message is broadcast, the trouble node and cluster state of the cluster file system is obtained.

Above-mentioned monitor node is the increased node in existing cluster, node configuration cluster multicast address, energy Enough receive cluster broadcast message.By receiving the multicast message of each node in cluster file system, according to multicast The type of message and the message parameter corresponding with multicast message, on to multicast message in inquiry storage table Stored record knows the trouble node and cluster state of cluster file system, in shared memory cluster text Cast communication message analysis node running status is utilized in part system, is overcome in the prior art and be there is shared depositing In the peer-to-peer framework of storage formula cluster file system, cluster disposal ability, node failure lack statistical analysis and converge Problem and defect that total and collection swarm parameter can not be adjusted dynamically.

Specifically, the first acquisition module 2200 specifically for：Obtain the type of message of the multicast message；When When the type of message is application layer messages type, the first message parameter of the multicast message is obtained, this first disappears Breath parameter at least includes：The message numbering of the corresponding application layer messages of the multicast message, the multicast multicast message Node where the first ring the first ring label and the multicast multicast message node in first ring First sender address；When the type of message is that node adds type of message, the of the multicast message is obtained Two message parameters, second message parameter at least includes：The second ring where the node of the multicast multicast message The second ring label, second sender address and multicast of the node in second ring of the multicast multicast message Node member's list of the node self record of the multicast message.

After the first receiver module 2100 receives multicast message, the first acquisition module 2200 obtains the group The type of message of message is broadcast, type of message here is broadly divided into two classes, and a class is application layer messages type, One class is that node adds type of message.

When the multicast message is application layer messages type, the parameter of acquisition will at least include：The multicast message The ring label of the ring where the message numbering seq of corresponding application layer messages, the node of the multicast multicast message, That is address of the node of the first ring label ring_id1 of the first ring and the multicast multicast message in ring, i.e., The first sender address sender_id1 in one ring；When type of message is that node adds type of message, obtain The parameter for taking at least includes：The ring label of the ring where the node of the multicast multicast message, i.e. the of the second ring The address of the node in ring of two ring label ring_id2, the multicast multicast message, i.e. in the second ring second The member of the node member of the node self record of sender address sender_id2 and the multicast multicast message Set, i.e. node member's list proc_list.To the acquisition of above parameter in order to be given expression to multicast message Cluster state judge.

Wherein, as shown in figure 17, enquiry module 2300 is specifically included：First judging submodule 2310, First sub-module stored 2320, the second sub-module stored 2330.

Specifically, the first judging submodule 2310, for when the multicast message be application layer messages type when, According to the message numbering and the first ring label, judge to whether there is tool in the application layer messages table of the storage table There is the first stored record of the message numbering and the first ring label；First sub-module stored 2320, is used for When not existing first stored record in the application layer messages table, the first message ginseng of the multicast message is stored Count to the application layer messages table；Second sub-module stored 2330, for existing when in the application layer messages table During first stored record, judged result is corresponding with first sender address node in first ring A upper node there is information drop-out, show on this that node is doubtful trouble node, and store this and doubtful ask Inscribe the doubtful trouble node table in parameter to storage table of node.

The first judging submodule 2310, the first sub-module stored specifically included in above-mentioned enquiry module 2300 2320th, the second sub-module stored 2330, realizes when multicast message is application layer messages type, leads to The parameter using the multicast message for obtaining is crossed to the application layer messages table in storage table, in doubtful trouble node table Data with existing make multilevel iudge, corresponding parameter stored according to correlated results finally or added up or The corresponding operating of deletion, reaches the real-time collecting to cluster file system status information and monitoring.

Further, the second sub-module stored 2330 is specifically included：Acquiring unit 2331, judging unit 2331, Recording unit 2333, memory cell 2334.

Wherein, acquiring unit 2331, for according to the first sender address, obtaining the doubtful trouble node First node address in first ring；Judging unit 2331, for according to the first node address, The message numbering and the first ring label, judge to whether there is with the first segment in the doubtful trouble node table Second stored record of dot address, the message numbering and the first ring label；Recording unit 2333, is used for When there is second stored record in the doubtful trouble node table, increase the message note of the doubtful trouble node Record number of times；Memory cell 2334, for when in the doubtful trouble node table do not exist second stored record When, storage includes that first parameter of the first node address, the message numbering and the first ring label is doubted to this Like trouble node table.

Further, enquiry module 2300 is specifically included：Storage of second judging submodule the 2340, the 3rd Module 2350, the 3rd judging submodule 2360, the 4th judging submodule 2370, record sub module 2380, 4th sub-module stored 2390, the 5th judging submodule 23100, deletion submodule 23110.

Wherein, the second judging submodule 2340, for when the multicast message be node add type of message when, According to second ring mark and second sender address, in judging that node in the storage table adds message table With the presence or absence of the 3rd stored record with the second ring label and second sender address；3rd storage Module 2350, for when the node is added and do not exist three stored records in message table, storing the group The second message parameter to the node for broadcasting message adds message table；3rd judging submodule 2360, for working as When there is three stored records in node addition message table, according to node member's list, judge and be somebody's turn to do 3rd stored record is compared, with the presence or absence of the node for increasing or decreasing in node member's list of the multicast message Member；4th judging submodule 2370, for when in node member's list of the multicast message exist reduce Node member when, the Section Point address of the node member of the reduction is obtained, according to the Section Point address And second sender address, judge leaving in node table with the presence or absence of having the Section Point for the storage table Address and the 4th stored record of second sender address；Record sub module 2380, for being left when this When there is four stored records in node table, increase the Message Record number of times of the node member of the reduction；The Four sub-module storeds 2390, for when this leaves and does not exist four storage informations in node table, storing The second parameter including the Section Point address, second sender address leaves node table to this；5th sentences Disconnected submodule 23100, for when there is increased node member in node member's list of the multicast message, The 3rd node address of the increased node member is obtained, according to the 3rd node address and second sender Address, judges that this whether there is with the 3rd node address and second sender address in leaving node table The 5th stored record；Delete submodule 23110, for leaving node table when this in exist the 5th storage During record, the 5th stored record is deleted.

The second judging submodule 2340 specifically included in above-mentioned enquiry module 2300 to delete submodule 23110 realize when multicast message be node add type of message when, by using obtain multicast message Parameter message table is added to the node in storage table, the data with existing in node table is left and is made multilevel iudge, The final corresponding operating for being stored to corresponding parameter according to correlated results or being added up or deleted, reaches to collection The real-time collecting of group's file system state information and monitoring.

Specifically, when multicast message is application layer messages type, the second acquisition module 2400 is specifically included： 6th judging submodule 2410, the 7th judging submodule 2420.

Wherein, the 6th judging submodule 2410, for the doubtful trouble node table in, when the doubtful problem When Message Record number of times is there are in node reaching the especially doubtful trouble node of preset maximum value, judge that this is pre- If whether maximum is identical with the Message Record number of times of other nodes in addition to the especially doubtful trouble node, when When preset maximum value is identical with the Message Record number of times of other nodes in addition to the especially doubtful trouble node, sentence The cluster file system break for group service busy state；7th judging submodule 2420, for when default When maximum is different from the Message Record number of times of other nodes in addition to the especially doubtful trouble node, judging should Especially doubtful trouble node is the trouble node.

Correspondingly, when multicast message is that node adds type of message, the second acquisition module 2400, specifically Including：8th judging submodule 2430, the 9th judging submodule 2440, the tenth judging submodule 2450, 11st judging submodule 2460.

Wherein, the 8th judging submodule 2430, for judging that this leaves whether node table is empty；9th sentences , for not being sky when this leaves node table, and during this leaves node table, there is multiple in disconnected submodule 2440 During stored record with the identical Section Point address, judge and the identical Section Point address phase Corresponding node is the trouble node；Tenth judging submodule 2450, for being sky when this leaves node table, Judge in node addition message table, the storage with second sender address minimum in second ring is remembered Whether the quantity of record reaches preset value；11st judging submodule 2460, for adding message when the node In table, the quantity of the stored record with second sender address minimum in second ring reaches preset value When, judge that the cluster file system is token time-out frequency.

Above-mentioned each judge module and each judging submodule, enter in the type according to multicast message in storage table In the different inquiry of row and deterministic processes, by the ratio between the stored counts of longitudinal node in itself and horizontal node Compared with, judge cluster state and trouble node, timely and effectively the state of cluster file system is made detection with Judge.

Further, monitor node also includes：

3rd acquisition module 2500, for obtaining corresponding adjustment ginseng according to the trouble node and cluster state Number.

Multicast module 2600, for by the adjusting parameter multicast to the clustered node so that the clustered node According to current itself configuration of adjusting parameter adjustment.

Correspondingly, the 3rd acquisition module 2500 specifically for：

It is default according to first when the cluster file system residing for the clustered node is group service busy state Multiplying power tune up current message transmission window value for a new information transmission window value, according to the second default multiplying power Reduce each clustered node current maximum can transmission information value be that a new maximum can transmission information value；When When cluster file system residing for the clustered node is token time-out frequency, according to the 3rd default multiplying power The token time-out time for tuning up the clustered node is a new token time-out time.

Monitor node of the increase that the present invention is provided in cluster file system, overcomes and exists in the prior art In the peer-to-peer framework of shared memory cluster file system, cluster disposal ability, node failure lack system Meter analysis collects and collects problem and defect that swarm parameter can not be adjusted dynamically, realizes and cluster file system is asked Timely detection, discovery and the solution of topic.

Each embodiment in this specification is described by the way of progressive, what each embodiment was stressed All be the difference with other embodiment, between each embodiment identical similar part mutually referring to.

Although having been described for the preferred embodiment of the embodiment of the present invention, those skilled in the art once obtain Cicada basic creative concept, then can make other change and modification to these embodiments.So, it is appended Claim is intended to be construed to include preferred embodiment and falls into being had altered for range of embodiment of the invention And modification.

Finally, in addition it is also necessary to explanation, herein, such as first and second or the like relational terms are only Only be used for by an entity or operation with another entity or operate make a distinction, and not necessarily require or Imply between these entities or operation there is any this actual relation or order.And, term " bag Include ", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion so that including one The process of list of elements, method, article or terminal device not only include those key elements, but also including not having There are other key elements being expressly recited, or also include being this process, method, article or terminal device Intrinsic key element.In the absence of more restrictions, by wanting that sentence "including a ..." is limited Element, it is not excluded that also exist in addition in the process including the key element, method, article or terminal device Identical element.

Above-described is the preferred embodiment of the present invention, it should be pointed out that for the ordinary people of the art For member, some improvements and modifications can also be made under the premise of principle of the present invention is not departed from, these Improvements and modifications are also within the scope of the present invention.

Claims

1. a kind of monitoring method of shared memory cluster file system node communication, is applied to shared storage Monitor node in formula cluster file system, it is characterised in that including：

Receive the multicast message of clustered node in the cluster file system, the monitor node and the cluster Node is all located in the cluster file system；

Obtain the type of message and the message parameter corresponding with the type of message of the multicast message；

According to the message parameter, the stored record corresponding with the multicast message in inquiry storage table；

When a prefixed time interval is reached, deposited according to corresponding with the multicast message in the storage table Storage record, obtains the trouble node and cluster state of the cluster file system.

2. the monitoring method that shared memory cluster file system node according to claim 1 communicates, Characterized in that, the type of message for obtaining the multicast message and corresponding with the type of message disappearing Breath parameter, specifically includes：

Obtain the type of message of the multicast message；

When the type of message is application layer messages type, the first message parameter of the multicast message is obtained, The first message parameter at least includes：Message numbering, the group of the corresponding application layer messages of the multicast message First ring label of the first ring where broadcasting the node of the multicast message and the section of multicast message described in multicast First sender address of the point in first ring；

When the type of message is that node adds type of message, the second message ginseng of the multicast message is obtained Number, second message parameter at least includes：Of the second ring where the node of multicast message described in multicast Second sender address and multicast of the node in second ring of two ring labels, multicast message described in multicast Node member's list of the node self record of the multicast message.

3. the monitoring method that shared memory cluster file system node according to claim 2 communicates, Characterized in that, it is described according to the message parameter, it is corresponding with the multicast message in inquiry storage table Stored record, specifically includes：

When the multicast message is application layer messages type, according to the message numbering and the first ring mark Number, judge to whether there is with the message numbering and described first in the application layer messages table of the storage table First stored record of ring label；

When not existing first stored record in the application layer messages table, the multicast message is stored First message parameter is to the application layer messages table；

When there is first stored record in the application layer messages table, judged result is first ring In the upper node of the node corresponding with first sender address there is information drop-out, draw on described One node is doubtful trouble node, and stores doubting in parameter to the storage table of the doubtful trouble node Like trouble node table.

4. the monitoring method that shared memory cluster file system node according to claim 3 communicates, Characterized in that, described draw a upper node for doubtful trouble node, and store the doubtful problem section Doubtful trouble node table in parameter to the storage table of point, specifically includes：

According to first sender address, first of the doubtful trouble node in first ring is obtained Node address；

According to the first node address, the message numbering and the first ring label, judge described doubtful With the presence or absence of with the first node address, the message numbering and the first ring mark in trouble node table Number the second stored record；

When there is second stored record in the doubtful trouble node table, increase the doubtful problem section The Message Record number of times of point；

When not existing second stored record in the doubtful trouble node table, storage includes described first First parameter of node address, the message numbering and the first ring label is to the doubtful trouble node table.

5. the monitoring method that shared memory cluster file system node according to claim 2 communicates, Characterized in that, it is described according to the message parameter, it is corresponding with the multicast message in inquiry storage table Stored record, specifically includes：

When the multicast message is that node adds type of message, according to second ring mark and described second Sender address, judges that the node in the storage table whether there is with second ring in adding message table 3rd stored record of label and second sender address；

When the node is added does not exist three stored record in message table, the multicast message is stored The second message parameter to the node add message table；

When the node is added there is three stored record in message table, arranged according to the node member Table, judges compared with the 3rd stored record, to whether there is in node member's list of the multicast message The node member for increasing or decreasing；

When there is the node member for reducing in node member's list of the multicast message, the reduction is obtained Node member Section Point address, according to the Section Point address and second sender address, Judge leaving in node table with the presence or absence of having the Section Point address and described second for the storage table 4th stored record of sender address；

When it is described leave node table in there is four stored record when, increase the node member of the reduction Message Record number of times；

When it is described leave node table in the absence of four storage information when, storage include the Section Point Address, the second parameter of second sender address leave node table to described；

When there is increased node member in node member's list of the multicast message, the increase is obtained Node member the 3rd node address, according to the 3rd node address and second sender address, Judge described leaving in node table with the presence or absence of there is the 3rd node address and second sender 5th stored record of location；

When it is described leave node table in there is five stored record when, delete the 5th stored record.

6. the monitoring method that shared memory cluster file system node according to claim 3 communicates, Characterized in that, it is described when a prefixed time interval is reached, disappear with the multicast according in the storage table The corresponding stored record of manner of breathing, obtains the trouble node and cluster state of the cluster file system, specific bag Include：

In the doubtful trouble node table, reached when there are Message Record number of times in the doubtful trouble node During the especially doubtful trouble node of preset maximum value, judge that the preset maximum value especially doubtful is asked with except described Whether the Message Record number of times of other nodes outside topic node is identical；

When preset maximum value and the Message Record number of times of other nodes in addition to the especially doubtful trouble node When identical, judge that the cluster file system is group service busy state；

When preset maximum value and the Message Record number of times of other nodes in addition to the especially doubtful trouble node When different, judge that the especially doubtful trouble node is described problem node.

7. the monitoring method that shared memory cluster file system node according to claim 5 communicates, Characterized in that, it is described when a prefixed time interval is reached, disappear with the multicast according in the storage table The corresponding stored record of manner of breathing, obtains the trouble node and cluster state of the cluster file system, specific bag Include：

Leave whether node table is empty described in judging；

When described leave node table and be not sky, and it is described leave node table, existing multiple has identical institute When stating the stored record of Section Point address, judge corresponding with Section Point address described in the identical Node is described problem node；

When the node table that leaves is sky, in judging node addition message table, with second ring Whether the quantity of the stored record of minimum second sender address reaches preset value；

In the node adds message table, with second sender address minimum in second ring The quantity of stored record when reaching preset value, judge that the cluster file system is token time-out frequency.

8. the monitoring method that shared memory cluster file system node according to claim 1 communicates, Characterized in that, the monitoring method also includes：

Corresponding adjusting parameter is obtained according to described problem node and cluster state；

By the adjusting parameter multicast to the clustered node, so that the clustered node is according to the adjustment ginseng Current itself configuration of number adjustment.

9. the monitoring method that shared memory cluster file system node according to claim 8 communicates, It is described that corresponding adjusting parameter is obtained according to described problem node and cluster state, specifically include：

It is pre- according to first when the cluster file system residing for the clustered node is group service busy state If multiplying power tune up current message transmission window value for a new information transmission window value, according to second default times Rate reduce each clustered node current maximum can transmission information value be that a new maximum can transmission information Value；

It is pre- according to the 3rd when the cluster file system residing for the clustered node is token time-out frequency If multiplying power to tune up the token time-out time of the clustered node be a new token time-out time.

10. the monitor node in a kind of shared memory cluster file system, it is characterised in that including：

First receiver module, the multicast message for receiving clustered node in the cluster file system is described Monitor node is all located in the cluster file system with the clustered node；

First acquisition module, for obtaining the type of message of the multicast message and relative with the type of message The message parameter answered；

Enquiry module, it is corresponding with the multicast message in inquiry storage table for according to the message parameter Stored record；

Second acquisition module, for when a prefixed time interval is reached, according in the storage table with it is described The corresponding stored record of multicast message, obtains the trouble node and cluster state of the cluster file system.

Monitor node in 11. shared memory cluster file systems according to claim 10, its Be characterised by, first acquisition module specifically for：

Obtain the type of message of the multicast message；

Monitor node in 12. shared memory cluster file systems according to claim 11, its It is characterised by, the enquiry module is specifically included：

First judging submodule, for when the multicast message is application layer messages type, being disappeared according to described Breath numbering and the first ring label, judge to whether there is with institute in the application layer messages table of the storage table State the first stored record of message numbering and the first ring label；

First sub-module stored, for when in the application layer messages table do not exist first stored record when, Store the first message parameter of the multicast message to the application layer messages table；

Second sub-module stored, for when there is first stored record in the application layer messages table, Judged result is that a upper node of corresponding with first sender address node in first ring is deposited In information drop-out, a upper node is drawn for doubtful trouble node, and store the doubtful trouble node Doubtful trouble node table in parameter to the storage table.

Monitor node in 13. shared memory cluster file systems according to claim 12, its It is characterised by, second sub-module stored is specifically included：

Acquiring unit, for according to first sender address, obtaining the doubtful trouble node described First node address in first ring；

Judging unit, for according to the first node address, the message numbering and the first ring label, Judge in the doubtful trouble node table with the presence or absence of have the first node address, the message numbering and Second stored record of the first ring label；

Recording unit, for when there is second stored record in the doubtful trouble node table, increasing The Message Record number of times of the doubtful trouble node；

Memory cell, for when not existing second stored record in the doubtful trouble node table, depositing Storage includes the first parameter of the first node address, the message numbering and the first ring label to described Doubtful trouble node table.

Monitor node in 14. shared memory cluster file systems according to claim 11, its It is characterised by, the enquiry module is specifically included：

Second judging submodule, for when the multicast message is that node adds type of message, according to described Second ring mark and second sender address, judge that the node in the storage table is in adding message table No the 3rd stored record existed with the second ring label and second sender address；

3rd sub-module stored, for when the node add message table in do not exist the 3rd stored record When, the second message parameter to the node for storing the multicast message adds message table；

3rd judging submodule, for when the node add message table in there is three stored record when, According to node member's list, judge compared with the 3rd stored record, the node of the multicast message With the presence or absence of the node member for increasing or decreasing in members list；

4th judging submodule, for when the node that there is reduction in node member's list of the multicast message During member, obtain the Section Point address of the node member of the reduction, according to the Section Point address and Second sender address, judges leaving in node table with the presence or absence of having described second for the storage table 4th stored record of node address and second sender address；

Record sub module, for when it is described leave node table in there is four stored record when, increase institute State the Message Record number of times of the node member of reduction；

4th sub-module stored, for when it is described leave node table in the absence of four storage information when, Storage includes that node is left in the Section Point address, the second parameter of second sender address to described Table；

5th judging submodule, for when there is increased node in node member's list of the multicast message During member, obtain the 3rd node address of the increased node member, according to the 3rd node address and Second sender address, judges that described being left whether there is with the 3rd node address in node table And the 5th stored record of second sender address；

Delete submodule, for when it is described leave node table in there is five stored record when, delete institute State the 5th stored record.

Monitor node in 15. shared memory cluster file systems according to claim 12, its It is characterised by that second acquisition module is specifically included：

6th judging submodule, for the doubtful trouble node table in, deposited when in the doubtful trouble node When there is Message Record number of times to reach the especially doubtful trouble node of preset maximum value, the default maximum is judged Whether value is identical with the Message Record number of times of other nodes in addition to the especially doubtful trouble node, when default When maximum is identical with the Message Record number of times of other nodes in addition to the especially doubtful trouble node, judge The cluster file system is group service busy state；

7th judging submodule, for when preset maximum value and other in addition to the especially doubtful trouble node When the Message Record number of times of node is different, judge that the especially doubtful trouble node is described problem node.

Monitor node in 16. shared memory cluster file systems according to claim 14, its It is characterised by that second acquisition module is specifically included：

8th judging submodule, for judging described to leave whether node table is empty；

9th judging submodule, for leave node table and be not sky when described, and it is described leave node table, Have described in identical during the stored record of Section Point address in the presence of multiple, judge with described in the identical The corresponding node in Section Point address is described problem node；

Tenth judging submodule, for being sky when the node table that leaves, judges that the node adds message table In, whether the quantity of the stored record with second sender address minimum in second ring reaches Preset value；

11st judging submodule, for when the node add message table in, with second ring most When the quantity of the stored record of small second sender address reaches preset value, the group document is judged System is token time-out frequency.

Monitor node in 17. shared memory cluster file systems according to claim 10, its It is characterised by, the monitor node also includes：

3rd acquisition module, for obtaining corresponding adjusting parameter according to described problem node and cluster state；

Multicast module, for by the adjusting parameter multicast to the clustered node, so that the clustered node According to current itself configuration of adjusting parameter adjustment.

Monitor node in 18. shared memory cluster file systems according to claim 17, institute The 3rd acquisition module is stated, specifically for：