CN101751415A - Metadata service system metadata synchronized method and writing server updating method - Google Patents
Metadata service system metadata synchronized method and writing server updating method Download PDFInfo
- Publication number
- CN101751415A CN101751415A CN200810224708A CN200810224708A CN101751415A CN 101751415 A CN101751415 A CN 101751415A CN 200810224708 A CN200810224708 A CN 200810224708A CN 200810224708 A CN200810224708 A CN 200810224708A CN 101751415 A CN101751415 A CN 101751415A
- Authority
- CN
- China
- Prior art keywords
- server
- reading
- writing
- metadata
- read
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a metadata service system metadata synchronized method and writing server updating method. The metadata service system provided by the invention includes a writing server and a reading server; the writing server is used for storing the metadata in a concurrent file system to receive reading access and receive writing access; correcting the metadata in the concurrent file system and synchronously updating the corrected metadata into the reading server; the reading server is used for storing the metadata and receiving reading access, receiving the synchronous updating of the writing server to the metadata, and is also used for converting into writing service when the write server is monitored to be valid. The invention solves the problem of single point failure through the mutual backup of the reading server and the writing server, and can meet high effective large-scale concurrent access demands.
Description
Technical field
The present invention relates to the distributed storage in the cluster calculating, relate in particular to the metadata service system that the metadata in the parallel file system is stored, revised and reads, and metadata synchronization method and writing server updating method.
Background technology
File system is to be used for the storage of data and the subsystem that reads in the computer system, generally is structured on the disk.In the network application scene, calculating separates often with storage, promptly by the distributed file system of access to netwoks.Network service for the high-performance calculation of using a large amount of computing machine concurrent workings and high capacity is used, and usually needs to use many servers that are exclusively used in storage that file system service is provided jointly, satisfies the access bandwidth demand.
A kind of relatively framework of high performance parallel file system commonly used is that " metadata " such as index informations of file system deposited in meta data server, with deposit data at other data servers, CLIENT PROGRAM only need be fetched a spot of index information from meta data server, in the time of major part, directly and data server mutual.This architectural configurations is simple, and higher access bandwidth can be provided, and is applicable to the great majority application.
But the meta data server of this framework is a single-point, and its inefficacy will cause whole file system unavailable, therefore has availability issue, and carrier grade service can't be provided.Simultaneously, meta data server is as a single-point, and its performance may become performance bottleneck when visiting in the face of highdensity peak value.
In the parallel file system framework of present existing multivariate data server, major part is the active/standby server that constitutes by shared storage device, finishing by means of shared storage device synchronously between the active/standby server, but when the generation active/standby server is switched, can lose partial status information.
In addition, some parallel file system visits the index information that is positioned on the multiple servers by the multi-hop inquiry of relay in the prior art, as PVFS2, but this can cause inquiring about the increase of time delay and be difficult to guarantee in the file system consistance of information on the multiple servers.
In sum, in the prior art, there is following shortcoming for the metadata management in the parallel file system:
1, the existing parallel file system that provides metadata to serve by Single-Server does not possess high availability, and this meta data server lost efficacy and will cause whole file system unavailable.
2, use shared storage device to constitute the meta data server of active and standby structure, when switching, can lose status data.And synchronization has only a station server that service is provided, and when towards a large amount of concurrent visit, is difficult to the load-bearing capacity that provides enough.
3, by multinode distributed storage metadata, promptly the mode by mutual Query Information provides a plurality of meta data server service manner, can influence the efficient of visit, increases the time delay of visit.Do not solve simultaneously the stationary problem of status data yet.
Summary of the invention
The invention provides a kind of metadata service system, solve the single point failure problem, and can satisfy high efficiency a large amount of concurrent requirements for access.
According to metadata service system provided by the invention, the present invention also provides a kind of metadata synchronization method, realizes the synchronous renewal of the metadata of storing between each server, guarantees the consistance of each server stores data.
According to metadata service system provided by the invention, the present invention also provides a kind of writing server updating method, when writing server lost efficacy, is converted to writing server by reading server, solves the single point failure problem.
Metadata service system provided by the invention comprises: writing server and read server;
Described writing server is used for the metadata of memory parallel file system, accepts read access; And accept write access, revise described metadata, and amended metadata synchronization is updated to the described server of reading;
The described server of reading is used to store described metadata, accepts read access; And accept of the synchronous renewal of described writing server to metadata; Whether also be used to monitor described writing server and lost efficacy, when monitoring described writing server when losing efficacy, the described server of reading is converted to described writing server.
Metadata service system provided by the invention comprises that at least two are read server, also comprises the mediation service device;
Described mediation service device is used for measuring and being activated when described writing server lost efficacy at the described server of reading, and determines described at least one that read in the server and is converted to described writing server.
The correspondence that is converted to described writing server is read server, also is used for the metadata synchronization of Last modification is read server to all the other.
Described mediation service device is an entity apparatus that is independent of described writing server and reads server; Perhaps
Described mediation service device is integrated in described reading in the server.
When described mediation service device is to be independent of described writing server and when reading entity apparatus of server, also to be used to store described writing server and to read server info; Receive the server info query requests that client is initiated, return described writing server and read server info.
Metadata synchronization method provided by the invention is applied to metadata service system provided by the invention, comprising:
Described writing server is accepted write request, and the metadata of this modification is write the temporary realm, and sends the write request announcement to the described server of reading;
Described read server and receive the announcement of described write request after, to the corresponding element data quiescing before revising, perhaps wait for the described corresponding element data read that is the reading back quiescing that finishes; And behind quiesce-completed, return announcement reception response to described writing server;
Described writing server receives described reading and the described metadata of storing in the described temporary realm is sent to the described server of reading after announcement that server returns receives response; And, upgrade local corresponding element data recording of storing with the described metadata of storing in the described temporary realm;
The described corresponding element data recording of reading the local storage of server update, and upgrading successfully back releasing quiescing.
Also comprise:
Whether the described server of reading of described writing server monitoring lost efficacy;
Described writing server only sends described write request announcement to the current server of reading that did not lose efficacy; And after determining current reading server and all returning announcement and receive response of not losing efficacy, with the described metadata of storing in the described temporary realm send to current do not lose efficacy read server.
If the described server of reading monitors described writing server inefficacy behind quiesce-completed, then remove quiescing.
Writing server updating method provided by the invention is applied to metadata service system provided by the invention, comprising:
Describedly read the current state that server is monitored described writing server;
When the described server of reading monitors described writing server when losing efficacy, the described server of reading is converted to described writing server.
Also comprise the mediation service device in described metadata service system, the described server of reading disposes two at least, and the described server of reading monitors described writing server when losing efficacy, and specifically comprises:
The described server of reading is initiated requests for arbitration to described mediation service device;
Described mediation service device is ranked and inherits the order of succession of writing server for respectively reading server;
According to described order of succession, have the maximum described server of reading that upgrades sequence number and be converted to described writing server; Wherein, described renewal sequence number is the corresponding sequence number that increases progressively according to the order of sequence of the described writing server metadata of distributing to each renewal, and amended metadata synchronization is being updated to the described described renewal sequence number that carries when reading server.
Above-mentioned writing server updating method specifically comprises:
When the described server of reading monitors described writing server when losing efficacy, initiate the requests for arbitration that request locks to described mediation service device, carry the up-to-date renewal sequence number of local storage;
The up-to-date renewal sequence number of server is respectively read in the storage of described mediation service device, is ranked and inherits the order of succession of writing server for respectively reading server, and distribute lock to give the server of reading of the first order of succession correspondence;
What obtain locking reads server according to the up-to-date renewal sequence number of storing in the described mediation service device of respectively reading server, judges whether the up-to-date renewal sequence number of local storage is wherein maximum renewal sequence number; If then self is converted to described writing server; Otherwise, lock abdicated to the next one in the described order of succession reads server.
When described mediation service device is integrated in describedly when reading in the server, set in advance and arbitraryly describedly read described mediation service device integrated in the server and be the mediation service device of having the right, receive described requests for arbitration by the described mediation service device of having the right;
When reading server under the described mediation service device of having the right and be converted to writing server, by read under this server select at random one all the other describedly read described mediation service device integrated in the server and be the mediation service device of having the right, and be notified to and respectively read server.
The present invention is by writing server and read server and form metadata service system.Writing server can not only be accepted read access, can also accept write access, revises the metadata in the parallel file system, and amended metadata synchronization is updated to reads server; Read the server stores metadata, accept read access; And accept the synchronous renewal of writing server to metadata; Monitor writing server when losing efficacy when reading server, be converted to writing server.Owing to be provided with mutually redundant server and the writing server read in the metadata service system provided by the invention, therefore, efficiently solve the single point failure problem.By data sync, respectively read the metadata of server stores unanimity, can backup each other each other, can satisfy a large amount of concurrent requirements for access; And be provided with because each server is parallel, do not need the multi-hop inquiry, every read access of reading server can direct reception client end, reading efficiency height.
Description of drawings
One of metadata service system structural representation that Fig. 1 provides for the embodiment of the invention;
Two of the metadata service system structural representation that Fig. 2 provides for the embodiment of the invention;
Three of the metadata service system structural representation that Fig. 3 provides for the embodiment of the invention;
The metadata synchronization update method process flow diagram that Fig. 4 provides for the embodiment of the invention;
One of writing server updating method process flow diagram that Fig. 5 provides for the embodiment of the invention;
Two of the writing server updating method process flow diagram that Fig. 6 provides for the embodiment of the invention;
Fig. 7 is an instantiation signaling process figure of writing server updating method.
Embodiment
The embodiment of the invention provides a kind of metadata service system, metadata synchronization update method and writing server updating method, backup each other each other by multiple servers, the synchronous renewal of the metadata of storing between each server and writing server upgrade, solve the single point failure problem, and can satisfy high efficiency a large amount of concurrent requirements for access.
Closed accompanying drawing below, System and method for provided by the invention is described in detail.
Referring to Fig. 1, one of metadata service system structural representation that provides for the embodiment of the invention comprises: writing server 11 and read server 12; Wherein:
Writing server 11 is used for the metadata of memory parallel file system, accepts read access; And accept write access, revise the metadata in the parallel file system, and amended metadata synchronization is updated to reads server 12;
Read server 12, be used for the metadata of memory parallel file system, accept read access; And the synchronous renewal of accepting 11 pairs of metadata of writing server; Also be used to monitor writing server 11 and whether lost efficacy, when monitoring writing server 11 inefficacies, read server 12 and be converted to writing server.
In system shown in Figure 1, disposed a writing server and one and read server.Writing server and read server and can accept read access; Writing server can be realized the modification and the data sync of metadata; When this writing server lost efficacy, read server and be converted to writing server, accept read access and write access, solve the single point failure problem.
Among one embodiment, metadata service system also comprises mediation service device 13, and reads server 12 and dispose two at least, and its structural representation as shown in Figure 2, wherein, mediation service device 13 is an entity apparatus that is independent of writing server 11 and reads server 12.Each component function is as follows:
Read server 12, be used for storing metadata, accept read access; And the synchronous renewal of accepting 11 pairs of metadata of writing server; Also be used to monitor writing server 11 and whether lost efficacy, when monitoring writing server 11 inefficacies, start mediation service device 13;
Among one embodiment, mediation service device 13 can also be stored writing server 11 and read the relevant information of server 12; Receive the server info query requests that client is initiated, return writing server and read server info.Client can be initiated read request to arbitrary server of reading, and initiate write request to writing server according to the server info that returns.
In the practical application, writing server can be the server with identical function with reading server.When a certain server is used as writing server, need to carry out the correlation function of writing server, that is: except the metadata in the memory parallel file system, accept outside the read access, also accept write access, revise the metadata in the parallel file system, and amended metadata synchronization is updated to respectively reads server.When a certain server when reading server, storing metadata is also accepted read access.All possess the writing server correlation function as each server of reading server, only do not open or temporary transient idle at this correlation function when reading server.When the writing server in the system lost efficacy, any one reads server can be converted to writing server.
Referring to Fig. 3, another structural representation of metadata service system for the embodiment of the invention provides comprises: writing server 11, at least two are read server 12 and mediation service device 13.Wherein, different with structure shown in Figure 2 is that mediation service device 13 is not an independent entity apparatus that is provided with, and reads in the server but be integrated in respectively.Being integrated in the mediation service device of respectively reading in the server 13 can be pure hardware module, also can be the corresponding function module of software and hardware combination.
The metadata service system that provides below in conjunction with the above embodiment of the present invention, to metadata synchronously more new technological process and the writing server when writing server lost efficacy more new technological process be specifically described.
Referring to Fig. 4, be metadata synchronization update method process flow diagram, specifically comprise:
Step S401, writing server are accepted write request, and the metadata of this modification is write the temporary realm;
Step S402, writing server send the write request announcement to reading server;
Step S403, read server and receive write request announcement after, to the corresponding element data quiescing before revising, perhaps wait for the corresponding element data read that is the reading back quiescing that finishes;
Return announcement to writing server behind step S404, the quiesce-completed and receive response;
Step S405, writing server receive to be read after announcement that server returns receives response, the metadata packing of storing in the temporary realm is sent to read server; And, upgrade local corresponding element data recording of storing with the metadata of storing in the temporary realm;
Step S406, read the corresponding element data recording of the local storage of server update, and is upgrading successfully releasing quiescing afterwards.
Above-mentioned steps 402 can promptly begin to carry out after writing server is accepted write request, promptly the metadata in this modification is written in the process of temporary realm, can send the write request announcement to reading server, needn't wait for that whole metadata of this modification are written into the execution in step S402 again that finishes.
In step S403, the corresponding element data are under an embargo and read; In order to avoid carry out read operation simultaneously in the follow-up execution data updating process, the feasible error in data that reads.
In step S406, read corresponding element data recording that server successfully upgrades local storage after, in time remove quiescing, accept the metadata read access again.When disposing more than one when reading server in the metadata service system, even other renewals of reading server are arranged not to be finished as yet, but the corresponding element data of reading server that these play pendulum do not allow to visit, so, do not have client process and have access to inconsistent metadata, thereby guaranteed the amended consistance of file system metadata.
Among one embodiment, writing server and respectively read server and monitor each other state mutually, and can circulate a notice of the status information that monitors mutually.In concrete the application, can be by at writing server with respectively read health monitors of operation (specifically monitoring method is a prior art, is not described further at this) in the server.Before reading the announcement of server transmission write request, writing server is determined and whether is read the server inefficacy according to current monitoring result at writing server; Only send the write request announcement to the current server of reading that did not lose efficacy.Writing server is read announcement that server returns when receiving response in reception, according to current monitoring result, after determining current respectively reading server and all having returned announcement and receive response of not losing efficacy, what just the metadata packing of storing in the temporary realm is sent to current not inefficacy respectively reads server.
By writing server with respectively read server and monitor each other state mutually, can guarantee that the synchronous renewal of metadata is carried out smoothly, specifically comprise:
1, if reading server lost efficacy in the quiescing process, after writing server is known, needn't wait for that the server of reading of this inefficacy returns announcement reception response, can proceed metadata synchronization to all the other servers of reading that did not lose efficacy.
2, if read server in the quiescing process, monitor writing server and lost efficacy, read can circulate a notice of mutually between the server, and remove quiescing.This write operation failure, but can not allow file system play pendulum.
If 3 writing servers lost efficacy, so, received reading server and can finishing renewal of metadata after the renewal that packing sends in the process that packing data sends; If read the metadata after server does not receive renewal in addition, can give its (in aftermentioned writing server method for synchronous, specifically describing) again synchronously by follow-up writing server so, guarantee respectively to read the metadata synchronization of storing in the server.
4, metadata packing send to read server after, if a certain server of reading lost efficacy, can not influence other and read data in server and upgrade, and writing server can in time know the server of reading of inefficacy, needn't wait for that this reads the server return message and just can judge whole data synchronization process end.
To sum up, under various possibility situations, metadata synchronization method provided by the invention can keep the consistance of metadata in the file system.Simultaneously, all read server can provide the read access service simultaneously, for a large amount of concurrent especially burst access very strong applicability is arranged.
According to the metadata service system that the above embodiment of the present invention provides, the present invention also provides a kind of writing server updating method, and its specific implementation flow process comprises as shown in Figure 5:
Step S501, read the current state of server monitoring writing server;
Step S502, judge whether writing server lost efficacy, if not, go to step S501; If writing server lost efficacy, execution in step S503;
Step S503, read server and be converted to writing server.
In metadata system, also comprise the mediation service device, and read server and dispose two at least, read server and monitor current writing server when losing efficacy that the idiographic flow that upgrades writing server comprises as shown in Figure 6:
Step S601, read server and initiate requests for arbitration, carry the up-to-date renewal sequence number of local storage to the mediation service device.
Wherein, upgrade sequence number and be the corresponding sequence number that increases progressively according to the order of sequence of the metadata that writing server distributes to each renewal, and carry this renewal sequence number when reading server in that amended metadata synchronization is updated to, respectively read server and preserve up-to-date renewal sequence number in this locality.
In the present embodiment, reading server and initiate requests for arbitration to the mediation service device, is the requests for arbitration that request locks.
The up-to-date renewal sequence number of server is respectively read in the storage of step S602, mediation service device, and is ranked and inherits the order of succession of writing server for respectively reading server.
The sort method of order of succession can be ranked by the sequencing of mediation service device according to the requests for arbitration that receives, and perhaps also can be ranked at random by the mediation service device.
Step S603, mediation service device distribute lock to give the server of reading of the first order of succession correspondence.
Step S604, the current server of reading that obtains locking compare the up-to-date renewal sequence number of storing in the up-to-date renewal sequence number of this locality storage and the mediation service device of respectively reading server.
Whether the up-to-date renewal sequence number of step S605, the local storage of judgement is wherein maximum renewal sequence number; If, execution in step S606; Otherwise, execution in step S607;
Step S606, the current server of reading that obtains locking are converted to writing server with self, process ends.
Step S607, the current server of reading that obtains locking are abdicated lock to the next one in the order of succession and are read server, go to step S604.
Referring to Fig. 7, be an instantiation signaling process figure of writing server updating method.Be described below:
Suppose to be provided with in the metadata service system and read server 1, read server 2, read server 3 and read server 4, reading server 1 is 1009 at the up-to-date renewal sequence number of this locality preservation, reading server 2 is 1008 at the up-to-date renewal sequence number of this locality preservation, reading server 3 is 1008 at the up-to-date renewal sequence number of this locality preservation, reading server 4 is 1009 at the up-to-date renewal sequence number of this locality preservation, according to aforementioned metadata synchronization method as can be known, respectively read the latest sequence number of storing in the server and at most only can differ 1.
Lost efficacy if monitor current writing server, and then needed to carry out writing server and upgrade, and reading server 1, read server 2, read server 3 and reading to determine in the server 4 server as writing server.Concrete signaling procedure is:
Read server 1, read server 2, read server 3 and read the requests for arbitration that server 4 locks to mediation service device initiation request respectively, and carry the up-to-date renewal sequence number of local storage;
Mediation service device storage reads server 1, read server 2, read server 3 and read the up-to-date renewal sequence number of server 4, and be ranked and inherit the order of succession of writing server for reading server 1, read server 2, read server 3 and reading server 4, be respectively: reading server 2 is first order of succession, reading server 4 is second order of succession, reading server 3 is the 3rd order of succession, and reading server 1 is the 4th order of succession; And distribute that lock gives the first order of succession correspondence read server 2;
The mediation service device can be notified to the order that is ranked and respectively read server, also the order that is ranked can be kept at this locality, initiatively knows by respectively reading server;
Read server 2 and know and oneself be first order of succession and obtained lock, read the up-to-date renewal sequence number of server, judge whether the up-to-date renewal sequence number of local storage is wherein maximum renewal sequence number according to all the other of storing in the mediation service device; Because reading the server 2 local up-to-date renewal sequence numbers of preserving is 1008, and the maximum of storing in mediation service device renewal sequence number is 1009, thereby the up-to-date renewal sequence number of judging local storage is not maximum renewal sequence number (also promptly not being up-to-date renewal sequence number, is that update times according to metadata increases progressively because upgrade sequence number); This is read server 2 and abandons lock, lock is abdicated to the next one in the order of succession (second order of succession) read server 4;
Read the comparison deterministic process that server 4 repeats to read server 2, because reading the server 4 local up-to-date renewal sequence numbers of preserving is 1009, and the maximum of storing in mediation service device renewal sequence number also is 1009, therefore, read server 4 and judge that the up-to-date renewal sequence number of local storage is the renewal sequence number of the maximum of storing in the mediation service device, to himself be converted to writing server (as new writing server), and write in the mediation service device;
Other the existing server (promptly reading server 4) of reading in the discovering server mediation service device is set to writing server, and then being provided with own is by synchronous server, keeps the server identity of reading of oneself;
The writing server of newly arbitrating out has maximum renewal sequence number, the renewal of its metadata is up-to-date, in order to make the metadata unanimity of respectively reading server stores, this writing server of newly arbitrating out (promptly reading server 4), to retransmit its last data updated and read server, carry out data synchronization updating to all the other; That is: read server 4 and retransmit its last data updated to being read server 1, read server 2 and reading server 3; Wherein, read server 2 and read server 3 to accept renewal, store up-to-date renewal sequence number 1009; The latest update sequence number of reading server 1 preservation has been 1009, has carried out corresponding renewal before showing, can ignore existing the renewal, accepts this renewal.
In the example shown in Figure 7, the mediation service device is an entity apparatus that is independent of writing server and reads server.When the mediation service device is integrated in when respectively reading in the server, can set in advance and arbitraryly read mediation service device integrated in the server and be the mediation service device of having the right, when writing server lost efficacy, signaling process is similar substantially, respectively reads server and initiates requests for arbitration to this mediation service device of having the right of setting.
Reading server under the mediation service device of having the right is converted to after the writing server, if lost efficacy, next time is when arbitrating new writing server, respectively read server and do not know that this initiates requests for arbitration to being integrated in the mediation service device which reads in the server, therefore, when reading server under the mediation service device of having the right and be converted to writing server, by read under this server select at random one all the other read mediation service device integrated in the server as next mediation service device of having the right, and be notified to and respectively read server.
One of ordinary skill in the art will appreciate that all or part of step that realizes in the foregoing description method is to instruct relevant hardware to finish by program, this program can be stored in the computer read/write memory medium, as: ROM/RAM, magnetic disc, CD etc.
In sum, the present invention reads server by writing server, at least two and the mediation service device is formed metadata service system.Read server owing to be provided with at least two in the metadata service system provided by the invention, and when writing server lost efficacy, can one of them be converted to writing server by reading server, therefore, efficiently solve the single point failure problem.By data sync, many metadata of reading the server stores unanimity can backup each other each other, can satisfy a large amount of concurrent requirements for access; And because many read that server is parallel to be provided with, do not need the multi-hop inquiry, every read access of reading server can direct reception client end, reading efficiency height.
Obviously, those skilled in the art can carry out various changes and modification to the present invention and not break away from the spirit and scope of the present invention.Like this, if of the present invention these are revised and modification belongs within the scope of claim of the present invention and equivalent technologies thereof, then the present invention also is intended to comprise these changes and modification interior.
Claims (12)
1. a metadata service system is characterized in that, comprising: writing server and read server;
Described writing server is used for the metadata of memory parallel file system, accepts read access; And accept write access, revise described metadata, and amended metadata synchronization is updated to the described server of reading;
The described server of reading is used to store described metadata, accepts read access; And accept of the synchronous renewal of described writing server to metadata; Whether also be used to monitor described writing server and lost efficacy, when monitoring described writing server when losing efficacy, the described server of reading is converted to described writing server.
2. the system as claimed in claim 1 is characterized in that, described system comprises that at least two are read server, and described system also comprises the mediation service device;
Described mediation service device is used for monitoring and being activated when described writing server lost efficacy at the described server of reading, and determines described at least one that read in the server and is converted to described writing server.
3. system as claimed in claim 2 is characterized in that the correspondence that is converted to described writing server is read server, also is used for the metadata synchronization of Last modification is read server to all the other.
4. system as claimed in claim 3 is characterized in that, described mediation service device is an entity apparatus that is independent of described writing server and reads server; Perhaps
Described mediation service device is integrated in described reading in the server.
5. system as claimed in claim 4 is characterized in that, when described mediation service device is to be independent of described writing server and when reading entity apparatus of server, also to be used to store described writing server and to read server info; Receive the server info query requests that client is initiated, return described writing server and read server info.
6. a metadata synchronization method is applied to the described metadata service system of claim 1, it is characterized in that, comprising:
Described writing server is accepted write request, and the metadata of this modification is write the temporary realm, and sends the write request announcement to the described server of reading;
Described read server and receive the announcement of described write request after, to the corresponding element data quiescing before revising, perhaps wait for the described corresponding element data read that is the reading back quiescing that finishes; And behind quiesce-completed, return announcement reception response to described writing server;
Described writing server receives described reading and the described metadata of storing in the described temporary realm is sent to the described server of reading after announcement that server returns receives response; And, upgrade local corresponding element data recording of storing with the described metadata of storing in the described temporary realm;
The described corresponding element data recording of reading the local storage of server update, and upgrading successfully back releasing quiescing.
7. method as claimed in claim 6 is characterized in that, also comprises:
Whether the described server of reading of described writing server monitoring lost efficacy;
Described writing server only sends described write request announcement to the current server of reading that did not lose efficacy; And after determining current reading server and all returning announcement and receive response of not losing efficacy, with the described metadata of storing in the described temporary realm send to current do not lose efficacy read server.
8. method as claimed in claim 7 is characterized in that, if the described server of reading monitors described writing server inefficacy behind quiesce-completed, then removes quiescing.
9. a writing server updating method is applied to the described metadata service system of claim 1, it is characterized in that, comprising:
Describedly read the current state that server is monitored described writing server;
When the described server of reading monitors described writing server when losing efficacy, the described server of reading is converted to described writing server.
10. method as claimed in claim 9, it is characterized in that, also comprise the mediation service device in described metadata service system, the described server of reading disposes two at least, and the described server of reading monitors described writing server when losing efficacy, and described method specifically comprises:
The described server of reading is initiated requests for arbitration to described mediation service device;
Described mediation service device is ranked and inherits the order of succession of writing server for respectively reading server;
According to described order of succession, have the maximum described server of reading that upgrades sequence number and be converted to described writing server; Wherein, described renewal sequence number is the corresponding sequence number that increases progressively according to the order of sequence of the described writing server metadata of distributing to each renewal, and amended metadata synchronization is being updated to the described described renewal sequence number that carries when reading server.
11. method as claimed in claim 10 is characterized in that, specifically comprises:
When the described server of reading monitors described writing server when losing efficacy, initiate the requests for arbitration that request locks to described mediation service device, carry the up-to-date renewal sequence number of local storage;
The up-to-date renewal sequence number of server is respectively read in the storage of described mediation service device, is ranked and inherits the order of succession of writing server for respectively reading server, and distribute lock to give the server of reading of the first order of succession correspondence;
What obtain locking reads server according to the up-to-date renewal sequence number of storing in the described mediation service device of respectively reading server, judges whether the up-to-date renewal sequence number of local storage is wherein maximum renewal sequence number; If then self is converted to described writing server; Otherwise, lock abdicated to the next one in the described order of succession reads server.
12. as claim 10 or 11 described methods, it is characterized in that, when described mediation service device is integrated in described when reading in the server, set in advance and arbitraryly describedly read described mediation service device integrated in the server and be the mediation service device of having the right, receive described requests for arbitration by the described mediation service device of having the right;
When reading server under the described mediation service device of having the right and be converted to writing server, by read under this server select at random one all the other describedly read described mediation service device integrated in the server and be the mediation service device of having the right, and be notified to and respectively read server.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810224708XA CN101751415B (en) | 2008-12-09 | 2008-12-09 | Metadata service system, metadata synchronized method and writing server updating method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810224708XA CN101751415B (en) | 2008-12-09 | 2008-12-09 | Metadata service system, metadata synchronized method and writing server updating method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101751415A true CN101751415A (en) | 2010-06-23 |
| CN101751415B CN101751415B (en) | 2012-03-28 |
Family
ID=42478406
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200810224708XA Active CN101751415B (en) | 2008-12-09 | 2008-12-09 | Metadata service system, metadata synchronized method and writing server updating method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101751415B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102694825A (en) * | 2011-03-22 | 2012-09-26 | 腾讯科技（深圳）有限公司 | Data processing method and data processing system |
| CN102780571A (en) * | 2011-05-11 | 2012-11-14 | 中兴通讯股份有限公司 | Main board and spare board switching processing method and system |
| CN103019875A (en) * | 2012-12-19 | 2013-04-03 | 北京世纪家天下科技发展有限公司 | Method and device for realizing double main reconstruction of database |
| CN103369051A (en) * | 2013-07-22 | 2013-10-23 | 中安消技术有限公司 | Data server cluster system and data synchronization method |
| CN103580891A (en) * | 2012-07-27 | 2014-02-12 | 腾讯科技（深圳）有限公司 | Data synchronization method and system and servers |
| CN104158898A (en) * | 2014-08-25 | 2014-11-19 | 曙光信息产业股份有限公司 | Updating method of file layout in distributed file system |
| CN104268097A (en) * | 2014-10-13 | 2015-01-07 | 浪潮(北京)电子信息产业有限公司 | Metadata processing method and system |
| WO2015000103A1 (en) * | 2013-07-01 | 2015-01-08 | Empire Technology Development Llc | System and method for data storage |
| CN105045938A (en) * | 2015-09-17 | 2015-11-11 | 浪潮(北京)电子信息产业有限公司 | Metadata concurrent access method and system |
| CN105468718A (en) * | 2015-11-18 | 2016-04-06 | 腾讯科技（深圳）有限公司 | Data consistency processing method, device and system |
| CN106603665A (en) * | 2016-12-16 | 2017-04-26 | 无锡华云数据技术服务有限公司 | Cloud platform continuous data synchronization method and cloud platform continuous data synchronization device |
| CN108829496A (en) * | 2018-05-29 | 2018-11-16 | 阿里巴巴集团控股有限公司 | A kind of service calling method, device and electronic equipment |
| CN111045870A (en) * | 2019-12-27 | 2020-04-21 | 北京浪潮数据技术有限公司 | Method, device and medium for saving and restoring metadata |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100587692C (en) * | 2007-01-26 | 2010-02-03 | 华中科技大学 | Method and system for promoting metadata service reliability |
| CN101247417B (en) * | 2008-03-07 | 2011-07-27 | 中国科学院计算技术研究所 | Double-layer metadata processing system and method |
-
2008
- 2008-12-09 CN CN200810224708XA patent/CN101751415B/en active Active
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102694825A (en) * | 2011-03-22 | 2012-09-26 | 腾讯科技（深圳）有限公司 | Data processing method and data processing system |
| CN102780571A (en) * | 2011-05-11 | 2012-11-14 | 中兴通讯股份有限公司 | Main board and spare board switching processing method and system |
| CN103580891A (en) * | 2012-07-27 | 2014-02-12 | 腾讯科技（深圳）有限公司 | Data synchronization method and system and servers |
| CN103019875A (en) * | 2012-12-19 | 2013-04-03 | 北京世纪家天下科技发展有限公司 | Method and device for realizing double main reconstruction of database |
| US9684672B2 (en) | 2013-07-01 | 2017-06-20 | Empire Technology Development Llc | System and method for data storage |
| WO2015000103A1 (en) * | 2013-07-01 | 2015-01-08 | Empire Technology Development Llc | System and method for data storage |
| CN103369051A (en) * | 2013-07-22 | 2013-10-23 | 中安消技术有限公司 | Data server cluster system and data synchronization method |
| CN103369051B (en) * | 2013-07-22 | 2016-04-27 | 中安消技术有限公司 | A kind of data server cluster system and method for data synchronization |
| CN104158898A (en) * | 2014-08-25 | 2014-11-19 | 曙光信息产业股份有限公司 | Updating method of file layout in distributed file system |
| CN104158898B (en) * | 2014-08-25 | 2018-01-19 | 曙光信息产业股份有限公司 | The update method of file layout in a kind of distributed file system |
| CN104268097A (en) * | 2014-10-13 | 2015-01-07 | 浪潮(北京)电子信息产业有限公司 | Metadata processing method and system |
| CN104268097B (en) * | 2014-10-13 | 2018-02-06 | 浪潮(北京)电子信息产业有限公司 | A kind of metadata processing method and system |
| CN105045938A (en) * | 2015-09-17 | 2015-11-11 | 浪潮(北京)电子信息产业有限公司 | Metadata concurrent access method and system |
| CN105468718A (en) * | 2015-11-18 | 2016-04-06 | 腾讯科技（深圳）有限公司 | Data consistency processing method, device and system |
| CN105468718B (en) * | 2015-11-18 | 2020-09-08 | 腾讯科技（深圳）有限公司 | Data consistency processing method, device and system |
| CN106603665A (en) * | 2016-12-16 | 2017-04-26 | 无锡华云数据技术服务有限公司 | Cloud platform continuous data synchronization method and cloud platform continuous data synchronization device |
| CN108829496A (en) * | 2018-05-29 | 2018-11-16 | 阿里巴巴集团控股有限公司 | A kind of service calling method, device and electronic equipment |
| CN111045870A (en) * | 2019-12-27 | 2020-04-21 | 北京浪潮数据技术有限公司 | Method, device and medium for saving and restoring metadata |
| CN111045870B (en) * | 2019-12-27 | 2022-06-10 | 北京浪潮数据技术有限公司 | Method, device and medium for saving and restoring metadata |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101751415B (en) | 2012-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101751415B (en) | Metadata service system, metadata synchronized method and writing server updating method | |
| CN105814544B (en) | System and method for supporting persistent partition recovery in a distributed data grid | |
| US8856091B2 (en) | Method and apparatus for sequencing transactions globally in distributed database cluster | |
| US20070061379A1 (en) | Method and apparatus for sequencing transactions globally in a distributed database cluster | |
| US9201747B2 (en) | Real time database system | |
| CN102025550A (en) | System and method for managing data in distributed cluster | |
| CN113268472B (en) | Distributed data storage system and method | |
| CN102148850A (en) | Cluster system and service processing method thereof | |
| CN107623703B (en) | Synchronization method, device and system for Global Transaction Identifier (GTID) | |
| CN112039970B (en) | Distributed business lock service method, server, system and storage medium | |
| CN113220795B (en) | Data processing method, device, equipment and medium based on distributed storage | |
| CN105069152B (en) | data processing method and device | |
| CN104750757B (en) | A kind of date storage method and equipment based on HBase | |
| CN115510156A (en) | Cloud native high-availability database service providing system and method | |
| CN101436209A (en) | Method and apparatus for synchronizing multiple databases | |
| CN109639773A (en) | A kind of the distributed data cluster control system and its method of dynamic construction | |
| CN104753987B (en) | A kind of distributed conversation management method and system | |
| CN111291062A (en) | Data synchronous writing method and device, computer equipment and storage medium | |
| CN103384882A (en) | Method of managing usage rights in a share group of servers | |
| CN101778131A (en) | Data synchronization system | |
| CN111382132A (en) | Medical image data cloud storage system | |
| CN101789963A (en) | Data synchronization system | |
| CN113094431A (en) | Read-write separation method and device and server | |
| CN117271583A (en) | System and method for optimizing big data query | |
| CN115562849A (en) | Cache data method and system based on high availability |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |