CN106681864A - Data chasing method, cluster and system - Google Patents
Data chasing method, cluster and system Download PDFInfo
- Publication number
- CN106681864A CN106681864A CN201611260222.2A CN201611260222A CN106681864A CN 106681864 A CN106681864 A CN 106681864A CN 201611260222 A CN201611260222 A CN 201611260222A CN 106681864 A CN106681864 A CN 106681864A
- Authority
- CN
- China
- Prior art keywords
- cluster
- log recording
- data
- asynchronous
- clusters
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/14—Error detection or correction of the data by redundancy in operation
- G06F11/1402—Saving, restoring, recovering or retrying
- G06F11/1471—Saving, restoring, recovering or retrying involving logging of persistent data for recovery
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Debugging And Monitoring (AREA)
Abstract
The invention discloses a data chasing method and a cluster. The data chasing method comprises the steps that after a target cluster of a cloud system is recovered to a normal state from a fault state, log records in fault duration are obtained from other clusters in the same cluster group; the log records of the same object in the acquired log records are combined; and the target cluster operates the object according to the combined log records. The scheme can increase efficiency of data chasing.
Description
Technical field
The present invention designs a calculating machine field, particularly with regard to a kind of data driven method and cluster.
Background technology
In public cloud scene, using object as the elementary cell of operation.If changing a data, its implementation
It is not that directly data are modified, but first deletes data place object, then uploads new object.
In multiple access region (access zone, the AZ) scene of public cloud, by way of synchronization replication, Ke Yishi
The data trnascription of existing same object is stored in multiple AZ.User can get this object from any one AZ.
During toward multiple AZ synchronization replication objects, there may be some AZ not successfully obtain the copy of object.AZ is caused to replicate
The possibility situation of copy failure is very more, and roughly, (memory space of such as AZ is or not the mistake that can be divided on I/O path
Foot, or afterwards the storage medium of AZ breaks down and cannot write new data) and (the cluster event in AZ of physical environment abnormal state
The states such as barrier, cluster isolation, cluster exception) two big class, all represent the ability that the AZ does not possess processing data.
If this AZ for when stored copies fail on an AZ, storing failure subsequently from other AZ can obtain this
Individual object, realizes the reparation of this node data, and finally on all AZ, copy reaches unanimously.It is this do not store successful AZ and
Successful AZ has been stored consistent process has been reached be also referred to as data driven.
Referring to Fig. 1 (A), Fig. 1 (B), Fig. 1 (C) and Fig. 1 (D), it is assumed that there are 3 access areas in publicly-owned cloud system, respectively
It is access area 1, access area 2 and access area 3, the change of data in this 4 Description of Drawings public cloud.
Such as Fig. 1 (A), it is assumed that when data 1 and metadata 1 (metadata 1 is the metadata of data 1) is write, access area 1
It is normal with access area 2, and access area 3 breaks down.So access area 1 and access area 2 can be successfully written data 1
With metadata 1, and access area 3 can not write successfully.Access area 1 and access area 2 are write after success, all this can be write
Operation is logged by.
As shown in Fig. 1 (B), it is assumed that now access area 3 recovers normal, and access area 3 is according to the access area of access area 1/
The daily record in domain 2, obtains data 1 and metadata 2 and preserves from other access areas.
If public cloud receives the request for data 1 being updated to data 2.So access area 1 and access area 2 can be deleted first
Except data 1 and metadata 1, data 2 and metadata 2 are then write.And increase by 2 in access area 1 and access area 2 newly
Daily record.New daily record respectively describes deletion data 1 and metadata 1, and write data 2 and metadata 2.
As shown in Fig. 1 (C), access area 3 deletes data 1 and metadata 1, so as to data storage 2 and metadata 2.And it is raw
Into daily record.
As shown in Fig. 1 (D), access area 3 obtains data 2 and metadata from other access areas (such as access area 1)
2, and preserve, and generate daily record.Access area 3 is write after data 2 and metadata 2, whole pursuit process is completed, in public cloud
The data of all access areas obtain synchronization.
We can see that and, for access area 3, in order to keep data syn-chronization with other access areas, perform
A large amount of operations, occupy a large amount of system resources of access area 3.
The content of the invention
A kind of first aspect, there is provided embodiment of data driven method, includes：The target cluster of cloud system from fault recovery just
Chang Hou, obtains the log recording between the target cluster age at failure from other clusters of same collection group；From what is obtained
In log recording, the log recording of same object is merged；The target cluster according to the log recording after union operation,
The object is operated.The program can improve the efficiency of data driven.
Write operation log recording and deletion behaviour in a kind of possible design of first aspect, to the identical data of the object
Make log recording, merge into and do not operate.The number of executions of Operation Log is reduced, data driven efficiency is improve.
In a kind of possible design of first aspect, a plurality of write operation log recording to the object is merged into one and is write
Operation log recording.Improve data driven efficiency.
In a kind of possible design of first aspect, a plurality of deletion action log recording to the object merges into one
Deletion action log recording.Improve data driven efficiency.
In a kind of possible design of first aspect, the identical data of the same object includes：With identical bucket name, object
The data of name and cryptographic Hash.This solution provides and how to determine an object, and how to confirm a section in an object
The scheme of data.
In a kind of possible design of first aspect, it is described from it is same collection group other clusters in obtain failure during
Log recording, specially：Log recording during obtaining failure from the main cluster of same collection group.Optionally, it is also possible to
From same collection group, the log recording during failure is obtained in other clusters in addition to main cluster.
In a kind of possible design of first aspect, the log recording is asynchronous log recording；In target cluster event
During barrier, the main cluster generates asynchronous daily record to the operation for performing, and the asynchronous log recording is stored in described asynchronous
Journal file；
In a kind of possible design of first aspect, it is described from it is same collection group other clusters in obtain failure during
Log recording, specially：The all log recordings in asynchronous journal file are obtained from the main cluster.This solution provides one
Plant the concrete scheme for obtaining log recording between age at failure.
In a kind of possible design of first aspect, after the target cluster recovery is normal, the main cluster stopped day
The will record write asynchronous journal file.The program may insure that all daily records in asynchronous journal file are precisely object set
Daily record during group's failure.
A kind of second aspect, there is provided embodiment of data driven cluster, collection group includes target cluster and other clusters, institute
State object set group to break down, the target cluster from fault recovery it is normal after, the target cluster is used for：From the cluster
The log recording during failure is obtained in other clusters of group；Log recording from the log recording for obtaining, to same object
Merge；According to the log recording after union operation, the object is operated.
The target cluster of second aspect can perform first aspect or each method that may be designed of first aspect, and possess
Corresponding technique effect.Here is no longer separately illustrated.
A kind of third aspect, there is provided cloud system, the cloud system includes multiple access area access zone, first party
Target cluster mentioned in the various possible design of face, second aspect or first, second aspect, described in different
Access area.
Description of the drawings
Fig. 1 (A) -1 (D) is memory node failure schematic diagram；
Fig. 2 is the embodiment topological diagram of cloud system of the present invention；
Fig. 3 is the flow chart of data driven method embodiment of the present invention.
Specific embodiment
Cloud system (such as public cloud, enterprise's cloud or private clound) includes multiple access areas (access zone, AZ),
Each AZ includes one or more cluster.The same composition of cluster provided by different AZ collects group.Each collection group includes
One master (leader) cluster, remaining cluster is from (follow) cluster.Each AZ can externally provide same address and supply cloud
Main frame outside system is accessed, such as IP address.Identical object is preserved in multiple AZ, forms multiple pairs of same object
This, can improve data reliability, and reading speed.Each AZ can have an ID, used as the numbering of AZ.The present invention is implemented
Example can apply to the storage of object or the storage of file.
Cluster in same collection group may come from all access areas in cloud system, it is also possible to come from cloud system
In part access area.Each cluster includes at least one node, and node is, for example, server or computer, can be stored
Data.
Fig. 2 is a kind of embodiment topological diagram of cloud system.The public cloud includes access area 11, the and of access area 12
Access area 13.Access area 11 includes cluster 1, cluster 2 and cluster 3；Access area 12 includes cluster 4, cluster 5 and cluster 6；
Access area 13 includes cluster 7, cluster 8 and cluster 9.Collection group 21 includes cluster 1, cluster 4 and cluster 7；Collection group 22 includes
Cluster 2, cluster 5 and cluster 8；Collection group 23 includes cluster 3, cluster 6 and cluster 9.Each collection group includes at least one node
(not shown).
It is the flow chart of data driven method embodiment of the present invention referring to Fig. 3.In case of no particular description, number
Performed by target cluster according to the operation of driven method embodiment, be that the node in target cluster is performed specifically, perform operation
Node be properly termed as destination node.
11st, in a Ge Ji groups, between target cluster age at failure, receive after operation requests positioned at the cluster of main AZ, press
Operated according to operation requests, and recorded the log recording of the operation requests.
It is also likely to be removal request that the operation requests are probably write request.
In a total cloud, main (leader) AZ can be included and from (follow) AZ.Below with the operation requests
As a example by being write request, the processing procedure of operation requests is introduced.Data are carried in write request, bucket name and object can also be carried
Name.
One kind is achieved in that data arrive first at the cluster positioned at main AZ.The complete number of targets of cluster-based storage of main AZ
According to rear, then the cluster that target data is sent in this cluster positioned at other AZ is stored；Or the cluster of main AZ receives target
After data, while oneself is stored, the cluster for data being issued other AZ of this collection group is stored.Both keep same
Each company-data identical mode belongs to the method for synchronization in one Ge Ji groups.
Each cluster is performing data writing operation or is reading data manipulation, all current operation can be recorded journal file
In, generate the log recording of a write operation/read operation.
Because cluster is made up of node, therefore, the operation write data or read data is specifically performed by node, with
And corresponding log recording is generated, and log recording is stored in journal file.After cluster receives write request, specifically by collecting
Which node in group can be determined storing by cluster by specific algorithm, for example each node data storage in turn, or
Person randomly chooses a node data storage.Memory node can also be specified by write request.If main cluster is by synchronous side
Formula sends data to other clusters, then corresponding log recording is in synchronization log file.
For convenience.Hereinafter the collection group that the target cluster is located is called object set group.In the target
In collection group, positioned at the cluster of main AZ main cluster is referred to as, remaining cluster is referred to as from cluster.
The state of each cluster can be safeguarded by state server in collection group.State server can Shi Ji groups
In a node, or a special server.One kind is likely achieved in that state server is periodically searched
Collect the heartbeat message of each cluster, if being not received by the heartbeat message of certain cluster beyond the scheduled time, then just this
The state of individual cluster is updated to malfunction.
Each cluster in collection group, can obtain the state of other clusters in collection group by state server.Institute
After stating target clustering fault, main cluster can not send data to the target cluster.Main cluster does not generate synchronization log, and
It is to generate asynchronous daily record, and asynchronous log recording in asynchronous journal file.After data syn-chronization is realized, node asynchronous day
The node that will file is located can delete asynchronous journal file.
12, target cluster recovery is normal.Failed server is updated to normal condition the state of target cluster.
Failed server safeguards the state of each cluster.After trouble shooting, state server can detect object set
Group recovers normal.The cluster state information that state server is safeguarded can be perceived by other clusters.Such as state server handle
The normal information of target cluster recovery is sent to other clusters.
13, target cluster obtain failure from other clusters during log recording.
For example, the log recording during obtaining failure from main cluster, these log recordings are the behaviour that main cluster was performed
Make, if target cluster performs like this one time, it is possible to realize consistent with the data of main cluster, that is, realize data and chase after
Catch up with.It can be that main cluster is actively sent to target cluster that target cluster obtains the mode of log recording；It can also be target cluster
Actively obtain from main cluster.Optionally, each cluster can generate log recording after data storage.Therefore, except by leading
Cluster provides log recording, and the target cluster can also obtain log recording from the cluster that other did not occur failure.
A kind of optional acquisition is achieved in that：Between target cluster age at failure, the main cluster is log recording to different
In step journal file.Therefore, this step obtains the asynchronous journal file and can obtain between target cluster age at failure from main cluster
Log recording.
Another optional scheme is that every log recording can record its generation time when generating.Therefore, according to institute
State the time period that object set group breaks down, it is possible to obtain generate log recording of the time between target cluster age at failure.
14, in the log recording for obtaining, the daily record to same object is merged.
Each log recording is for an object.Daily record can record the ID of object.
The strategy of merging at least includes：To same object, newly-built and deletion identical data is merged into and not operated.
For same object, a data are whether first write, this data is then deleted again；Still it is first deleted
In a data, identical data are then write again.This can mean that this data had both been not written into or had not deleted.Therefore, so
One group of inverse operation can merge and bury in oblivion, that is, do not perform.
Therefore, to the write operation daily record and deletion action daily record of same data, merge into and do not operate.Equivalent to two behaviour
Bury in oblivion as daily record, two daily records are not performed.
Optionally, to a plurality of write operation log recording of same object, a write operation log recording is merged into.
Optionally, to a plurality of deletion action log recording of same object, a deletion action log recording is merged into.
Logically, object can be stored in bucket.The bucket name of different buckets is different.Inside same bucket, different objects
Object name it is different.Record has bucket name and object name in log recording.Therefore, using bucket name+object name, one can be uniquely determined
Individual object.Data in object, can be made a distinction with data content；Or the mark (such as cryptographic Hash) of data content is carried out
Distinguish.Log recording can be with the cryptographic Hash of record data.With identical bucket name, object name and identical data cryptographic Hash two
Bar log recording, is the log recording of the identical data for same target.
15, unenforced log recording in journal file is operated.
After the union operation of step 13, remaining log recording is operated.In remaining operation：If deleting
Division operation, then perform deletion action；If write operation, then the number for needing write is obtained from main cluster or other clusters
According to then writing the target cluster.The process of acquisition can be active (actively from other clusters obtain), or passive
(other clusters are actively sent to the target cluster).
Additionally, the present invention also provides an optional embodiment.Before step 11, that is, target cluster age at failure
Between, if the no and main company-data of other clusters outside main cluster is synchronous (such as breaking down).In the prior art, to wait
Data driven is performed again after all clusters all recover normally.This way reduces the reliability of data.What the present invention was provided
Embodiment is：Whether no matter there is cluster to be in malfunction in collection group, recovered normal cluster is not waited, held
Row data driven.
Therefore, before target cluster recovery is normal, other recovered normal clusters first carry out data driven, without
Wait all cluster recoveries in whole collection group normal.On the one hand this scheme improves data reliability, on the other hand reduces
Overall time of data driven.
Above-mentioned steps are described to data syn-chronization.In object storage, except data itself, also it is referred to as " data
Data " metadata, metadata is also required to keep synchronous between cluster；Its storage, synchronous flow process and data are completely the same,
Therefore individually introduction is not done.
The present invention also carried a kind of data driven cluster, and (specifically, data driven cluster here refers to object set
Group) embodiment.Above-mentioned embodiment of the method can be performed.The relation of cluster and collection group may be referred to Fig. 2 and close above
In the introduction of Fig. 2.
Collection group includes target cluster and other clusters, and the object set group breaks down, in the target cluster from event
After barrier recovers normal, the target cluster is used for：Log recording during obtaining failure from other clusters of the collection group；
From the log recording for obtaining, the log recording of same object is merged；According to the log recording after union operation, to institute
State object to be operated.
A kind of optional mode of combining log records is：Log recording to the identical data of same object, write operation and
Deletion action is merged into and not operated.Judge whether it is that a kind of optional mode of the identical data of same object is：With identical bucket
The data of name, object name and cryptographic Hash are the identical datas of same object.Except cryptographic Hash, can also be with other label descriptions
One data, for example, replace cryptographic Hash using the feature field of data.
The optional mode of another of combining log records is：Log recording of writing to the different pieces of information of same object, closes
And write log recording for one.Multiple log recordings before merging are converted into the single log recording after merging.Benefit is only to hold
One log recording of row, it is possible to reach the effect for performing multiple log recordings.Without log recording is performed a plurality of times, without many
The secondary data obtained from main cluster are synchronized, therefore improve the efficiency of data driven.
It is described from it is described collection group other clusters in obtain failure during log recording, specially：From the cluster
The log recording during failure is obtained in the main cluster of group.
Log recording during failure has various acquisition methods, such as described log recording is asynchronous log recording；Institute
State between target cluster age at failure, the main cluster generates asynchronous daily record to the operation for performing, and the asynchronous log recording is protected
There is the asynchronous journal file.Like this, all daily records for recording in asynchronous journal file, are exactly target cluster age at failure
Between operation performed by main cluster daily record.Get the asynchronous journal file and just get corresponding daily record.In object set
Before group's failure, the log recording of each collection group records is all synchronization log, in being stored in synchronization log file.
Optionally, after asynchronous daily record is acquired, asynchronous daily record is deleted.
It is described from it is described collection group other clusters in obtain failure during log recording, can be specially：From described
The all log recordings in asynchronous journal file are obtained in other clusters of collection group.
After the target cluster recovery is normal, the main cluster stops log recording writing the asynchronous daily record text
Part.It is changed to generate synchronization log, writes synchronization log file.
With reference to the embodiments described herein description each example unit and algorithm steps, can with electronic hardware,
Or computer software and electronic hardware are implemented in combination in.These functions are performed with hardware or software mode actually, are taken
Certainly in the application-specific and design constraint of technical scheme.Professional and technical personnel can be used not each specific application
Described function is realized with method.
Claims (13)
1. a kind of data driven method, it is characterised in that include：
The target cluster of cloud system from fault recovery it is normal after, from it is same collection group other clusters in obtain in the target
Log recording during clustering fault；
From the log recording for obtaining, the log recording of same object is merged；
The target cluster is operated according to the log recording after union operation to the object.
2. data driven method according to claim 1, the merging includes：
In log recording to the identical data of the object, write operation and deletion action are merged into and not operated.
3. data driven method according to claim 1, the identical data of the same object includes：
Data with identical bucket name, object name and cryptographic Hash.
4. data driven method according to claim 1 and 2, described to obtain event from other clusters of same collection group
Log recording during barrier, specially：
Log recording during obtaining failure from the main cluster of same collection group.
5. data driven method according to claim 4, wherein：
The log recording is asynchronous log recording；Between the target cluster age at failure, operation of the main cluster to performing
Asynchronous daily record is generated, and the asynchronous log recording is stored in the asynchronous journal file；
It is described from it is same collection group other clusters in obtain failure during log recording, specially：From the main cluster
The middle all log recordings obtained in asynchronous journal file.
6. data driven method according to claim 5, wherein：
After the target cluster recovery is normal, the main cluster stops log recording writing the asynchronous journal file.
7. a kind of data driven cluster, collects group including target cluster and other clusters, and the object set group breaks down, in institute
State target cluster from fault recovery it is normal after, it is characterised in that the target cluster is used for：
Log recording during obtaining failure from other clusters of the collection group；
From the log recording for obtaining, the log recording of same object is merged；
According to the log recording after union operation, the object is operated.
8. data driven method according to claim 7, the merging includes：
In log recording to the identical data of the object, write operation and deletion action are merged into and not operated.
9. data driven cluster according to claim 7, the identical data of the same object includes：
Data with identical bucket name, object name and cryptographic Hash.
10. the data driven cluster according to claim 7 or 8, described to obtain event from other clusters of the collection group
Log recording during barrier, specially：
Log recording during obtaining failure from the main cluster of the collection group.
11. data driven clusters according to claim 10, wherein：
The log recording is asynchronous log recording；Between the target cluster age at failure, operation of the main cluster to performing
Asynchronous daily record is generated, and the asynchronous log recording is stored in the asynchronous journal file；
It is described from it is described collection group other clusters in obtain failure during log recording, specially：From the main cluster
Obtain all log recordings in asynchronous journal file.
12. data driven clusters according to claim 7, wherein：
After the target cluster recovery is normal, the main cluster stops log recording writing the asynchronous journal file.
A kind of 13. cloud systems, including multiple access area access zone, it is characterised in that wherein：
Target cluster in the arbitrary described collection group of claim 7-9, from the different access areas.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611260222.2A CN106681864A (en) | 2016-12-30 | 2016-12-30 | Data chasing method, cluster and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611260222.2A CN106681864A (en) | 2016-12-30 | 2016-12-30 | Data chasing method, cluster and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106681864A true CN106681864A (en) | 2017-05-17 |
Family
ID=58873460
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611260222.2A Pending CN106681864A (en) | 2016-12-30 | 2016-12-30 | Data chasing method, cluster and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106681864A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1180890A2 (en) * | 2000-08-17 | 2002-02-20 | FusionOne, Inc. | Change log aggregation and optimization |
| CN101727504A (en) * | 2010-01-29 | 2010-06-09 | 成都市华为赛门铁克科技有限公司 | Method and device for migrating data of file system |
| CN102831156A (en) * | 2012-06-29 | 2012-12-19 | 浙江大学 | Distributed transaction processing method on cloud computing platform |
| CN103870560A (en) * | 2014-03-06 | 2014-06-18 | 上海爱数软件有限公司 | Log managing method for cluster |
-
2016
- 2016-12-30 CN CN201611260222.2A patent/CN106681864A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1180890A2 (en) * | 2000-08-17 | 2002-02-20 | FusionOne, Inc. | Change log aggregation and optimization |
| CN101727504A (en) * | 2010-01-29 | 2010-06-09 | 成都市华为赛门铁克科技有限公司 | Method and device for migrating data of file system |
| CN102831156A (en) * | 2012-06-29 | 2012-12-19 | 浙江大学 | Distributed transaction processing method on cloud computing platform |
| CN103870560A (en) * | 2014-03-06 | 2014-06-18 | 上海爱数软件有限公司 | Log managing method for cluster |
Non-Patent Citations (1)
| Title |
|---|
| 张虎等: "一种基于日志合并优化的数据同步机制", 《小型微型计算机系统》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101259557B1 (en) | Cluster data management system and method for data recovery using parallel processing in cluster data management system | |
| CN104813321B (en) | The content and metadata of uncoupling in distributed objects store the ecosystem | |
| CN102024016B (en) | Rapid data restoration method for distributed file system (DFS) | |
| TWI222562B (en) | Dynamic links to file system snapshots | |
| CN103761165B (en) | Log backup method and device | |
| CN109522290B (en) | HBase data block recovery and data record extraction method | |
| CN100498796C (en) | Logic log generation method, database backup/ restoration method and system | |
| CN101576834B (en) | System and method for protecting continuous data for establishing data view based on time stamp | |
| CN106484565B (en) | Method of data synchronization and relevant device between multiple data centers | |
| US8527454B2 (en) | Data replication using a shared resource | |
| CA2913036A1 (en) | Index update pipeline | |
| US9525733B2 (en) | Method and system for data synchronization | |
| CN104156377A (en) | Virtual file system achieving reliability through multi-network-disk redundant storage | |
| WO2016061956A1 (en) | Data processing method for distributed file system and distributed file system | |
| CN104346373A (en) | Partition log queue synchronization management method and device | |
| CN102508740A (en) | Remote replication method of file system | |
| CN106681865B (en) | Service recovery method and device | |
| CN107046811A (en) | One introduces a collection storage device sends the clone file of source file and source file to the method, source storage device and backup storage device of backup storage device | |
| WO2017014814A1 (en) | Replicating memory volumes | |
| CN101937378B (en) | Method for carrying out back-up protection on data of storage equipment and computer system | |
| CN104572891B (en) | A kind of file updating method for network information separation storage | |
| CN108271420B (en) | Method for managing files, file system and server system | |
| CN108063782A (en) | Node is delayed machine adapting method and device, node group system | |
| CN116483284B (en) | Method, device, medium and electronic equipment for reading and writing virtual hard disk | |
| CN103544081A (en) | Management method and device for double metadata servers |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170517 |