CN112000347A

CN112000347A - Distributed file system and component version updating method thereof

Info

Publication number: CN112000347A
Application number: CN202010731380.1A
Authority: CN
Inventors: 金松
Original assignee: New H3C Big Data Technologies Co Ltd
Current assignee: New H3C Big Data Technologies Co Ltd
Priority date: 2020-07-27
Filing date: 2020-07-27
Publication date: 2020-11-27
Anticipated expiration: 2040-07-27
Also published as: CN112000347B

Abstract

The application provides a distributed file system and a component version updating method thereof, wherein in the method, an object storage device OSD component of the distributed file system receives a first increment synchronous message carrying an increment map and a non-verification indication mark; the object storage device component calculates an unverified object storage device graph based on the increment graph carried by the first increment synchronous message increment graph; the version of the object storage device map is updated.

Description

Distributed file system and component version updating method thereof

Technical Field

The application relates to a storage technology, in particular to a distributed file system and a component version updating method thereof.

Background

The distributed file system Ceph provides a highly reliable, scalable storage system. In a radar (Reliable, automatic, Distributed Object storage) cluster of a Ceph architecture, an OSD (Object storage Device) component is responsible for completing data storage and maintenance functions, and an MON (monitor) component is responsible for completing system state detection and maintenance. Once an application accesses the Ceph cluster to perform a write operation, data is stored in the OSD as an object. The OSD map is the most important type of cluster information maintained by the MON component. All OSD nodes report the state to the MON node, the MON component collects the information to generate a new version of OSD map, and the OSD and the MON, the OSD and the OSD share the OSD map in the cluster to the whole Ceph cluster through elementary diffusion propagation.

In the Ceph architecture, the updates of all the components are not synchronized, and after the version of the MON component is updated, the fields of the updated OSD map are increased, but after the OSD map update is received by the un-updated OSD component, the OSD map obtained by decoding according to the old version does not contain the newly added fields. The OSD map stores a lot of contents required by the cluster, and a special OSD map needs to be generated every time the contents in the cluster are changed. The new version OSD map and the un-updated OSD map carry different version numbers to distinguish whether the updating/updating is carried out. Updating the OSD map version updates the contents of the OSD map.

FIG. 1 is a diagram illustrating the synchronization status of the conventional OSD map under the Ceph architecture; the MON component MON.1 and part of the OSD components have been updated to a new version, the MON component running in the new version sends a synchronization message (i) to the non-updated OSD component carrying the encoded (incremental) map e2 and CRC (cyclic redundancy check) of the OSD map, the non-updated OSD component OSD.0 decodes to obtain the encoded (incremental) map e2 without including fields in the new version, so that the calculated OSD map and the CRC obtained based on the calculated OSD map are different from the CRC from the MON component MON.1, the OSD component stores the calculated OSD map to the disk, then sends a message requesting the OSD map to the MON component MON.1, executes a callback function according to the OSD map storing the disk, the OSD component OSD.0 is not the same version, but the time interval of the OSD map sent by the MON component MON.1 is longer, the OSD component OSD map receives the OSD map from other OSD components OSD map 1-n, the old version of the old version is still processed with a very different version number of the old system, until the OSD component OSD.0 receives the message with the full OSD map from MON component MON.1, the OSD component OSD.1 writes the OSD map into the hard disk, and then updates the new version. The version of the osd.1 of the OSD component is not updated for a long time, and the difference of the OSD map version number in Ceph can cause a long Peering process of PG information which is negotiated and consistent among a plurality of OSDs (destination groups) of a consistent hash virtual node where the OSD component osd.0 is located, and storage service is influenced.

Disclosure of Invention

The application aims to provide a distributed file system and a component version updating method thereof, which can reduce the influence on business caused by slow updating of the version number of an upgrade object storage device.

In order to achieve the above object, the present application provides a version updating method for a distributed file system component, where an OSD component of any object storage device of a distributed file system receives a first increment synchronization message carrying an increment map and a non-verification indication flag; any OSD assembly calculates a non-verified full-amount OSD map based on the increment map carried by the first increment synchronous message; and updating the full OSD map version.

In order to achieve the above object, the present application further provides a distributed file system, where the system includes a plurality of object storage device components and a plurality of monitor components, where any one object storage device OSD component of the distributed file system receives a first increment synchronization message carrying an increment map and a non-verification indication flag; any OSD component calculates the unverified full-amount OSD map based on the increment map carried by the first increment synchronous message increment map; and updating the full OSD map version.

In the application, the OSD component controls the OSD map not to execute CRC check in the online upgrading process according to the indication of the monitor serving as the primary equipment in the system, and eliminates the influence on the service caused by the slow version number updating of the online upgrading OSD map.

Drawings

FIG. 1 is a state diagram of object storage device graph synchronization under the Ceph architecture;

FIG. 2 is a schematic diagram illustrating a distributed file system component version update method provided herein;

FIG. 3 is a schematic diagram illustrating a selection of a full object storage device graph synchronization mode provided by the present application;

FIG. 4 is a schematic diagram illustrating an embodiment of a full volume object storage device graph synchronization schema provided herein;

fig. 5 is a schematic diagram of a second embodiment of a full-scale object storage device graph synchronization mode provided in the present application.

Detailed Description

A detailed description will be given of a number of examples shown in a number of figures. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the present application. Well-known methods, procedures, components and circuits have not been described in detail so as not to obscure the examples.

The term "including" as that term is used is meant to include, but is not limited to; the term "comprising" means including but not limited to; the terms "above," "within," and "below" include the instant numbers; the terms "greater than" and "less than" mean that the number is not included. The term "based on" means based on at least a portion thereof.

FIG. 2 is a schematic diagram illustrating a component version updating method of a distributed file system according to the present application; the method comprises the following steps:

step 101, an OSD component receives a first increment synchronous message carrying an increment map and a non-verification indication mark;

102, calculating an unverified full-quantity OSD map by the OSD component based on the increment map carried by the first increment synchronous message; and updating the full OSD map version.

In the application, the OSD component controls the OSD map not to execute CRC check in the online upgrading process according to the indication of the monitor serving as the primary equipment in the system, and eliminates the influence on the service caused by the slow updating of the version number of the full-scale OSD map for online upgrading.

Fig. 3 is a schematic diagram illustrating selection of a full-scale object storage device graph synchronization mode provided in the present application, where the embodiment includes:

in step 301, the Primary monitor component obtains a clustered version of Ceph.

In a plurality of monitor components under the Ceph framework, after a primary monitor serving as a primary device carries out full-scale OSD map, the primary monitor starts synchronizing the increment map under the Ceph framework.

Updating versions of all components under the Ceph framework are not synchronous, when only a small number of components under the Ceph framework update the whole OSD map, the cluster version of the Ceph is the version before updating; when most or all components under the Ceph architecture update the full amount of OSD map, the cluster version of the Ceph is an updated new version.

Step 302, judging whether the Primary monitor component is updated to a new version or not; if not, non-CRC check processing 310 is performed, and if yes, CRC check processing 320 is performed.

The primary monitor judges the version updating state of the components under the Ceph framework by acquiring the cluster version number, does not execute CRC check when a small part of the components update the full OSD map and the primary monitor synchronizes the increment map, and executes CRC check when the large part of the components or all the components update the full OSD map.

FIG. 4 is a schematic diagram of an embodiment of a full-scale object storage device graph synchronization mode provided in the present application, where the mode includes the following steps:

in step 311, the Primary monitor component sends an increment synchronization message carrying the increment map and the non-verification indication identifier.

In step 312, the OSD component receives the increment synchronization message carrying the increment map and the non-verification indication identifier carried by the increment map, calculates the non-verification full-size OSD map, and updates the non-verification full-size OSD map into a new version full-size OSD map.

Fig. 5 is a schematic diagram of a second embodiment of a full-scale object storage device graph synchronization mode provided in the present application, where the mode includes the following steps:

in step 321, the Primary monitor component sends an increment synchronization message carrying the increment map, the CRC value and the check indication identifier.

In step 322, the OSD component receives the incremental synchronization message carrying the increment map, the CRC value, and the check indication flag.

And 323, calculating the total OSD map to be checked by the OSD component according to the received increment map, and calculating the CRC value to be checked according to the total OSD map to be checked.

In step 324, the OSD component determines that the CRC to be checked is consistent with the received CRC? If yes, go to step 325; if not, go to step 326

Step 325, update to the full OSD map of the new version.

In step 326, the OSD component sends the full OSD map request message to the Primary Monitor without executing the callback function.

In step 327, the Primary monitor component sends a full synchronization message carrying a full OSD map.

In step 328, the OSD component receives the full-size OSD map request message and updates the full-size OSD map to the new version.

The OSD components do not call back the increment maps with inconsistent CRC any more, and directly request the full OSD maps, thereby avoiding the problems of long Peering process and long service interruption time caused by large difference of version numbers of the components of the Ceph.

The above description is only exemplary of the present application and should not be taken as limiting the present application, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A distributed file system component version updating method is characterized in that,

an object storage device OSD component of a distributed file system receives a first increment synchronous message carrying an increment map and a non-verification indication mark;

the OSD component calculates a non-verified full-amount OSD map based on the increment map carried by the first increment synchronous message; and updating the full OSD map version.

2. The method of claim 1, further comprising:

the OSD component receives a second increment synchronous message carrying the increment map, the Cyclic Redundancy Check (CRC) value and the check indication mark;

the OSD component calculates the total OSD map to be checked based on the increment map carried by the second increment synchronous message;

the OSD component calculates a CRC value to be checked according to the full OSD map to be checked;

the OSD component compares the CRC value of the second incremental synchronization message with the calculated CRC value to be checked;

the OSD component determines that the difference exists, does not execute the callback function process and sends a full synchronization request message to the monitor component which sends the second incremental synchronization message;

the monitor component sending a full synchronization message with the new version of the full OSD map to the OSD component;

and the OSD component receives the full synchronization message, acquires the full OSD map of the new version and updates the full OSD map of the new version.

3. The method of claim 2, further comprising:

the OSD component determines that the CRC value of the second increment synchronous message is consistent with the calculated CRC value to be checked, and executes a callback function process;

and the OSD component is updated to the full OSD map of the new version.

4. The method of claim 1, wherein before the OSD component of the object storage device of the distributed file system receives the first delta synchronization message carrying the delta map increment map and the non-check indication flag, the method further comprises:

the monitor component obtains a cluster version of the distributed file system;

the monitor component determines that the version number of the cluster version does not correspond to the version number of the new version full amount of OSD map;

the monitor component sends the first delta synchronization message.

5. The method of claim 1, wherein before the OSD component receives a second incremental synchronization message carrying the increment map, cyclic redundancy check, CRC, value, and check indication flag, the method further comprises:

the monitor component determines that the version number of the re-acquired cluster version corresponds to the version number of the new version full OSD map;

the monitor component sends the second delta synchronization message.

6. A distributed file system comprising a plurality of object storage device components and a plurality of monitor components,

any object storage device OSD component of the distributed file system receives a first increment synchronous message carrying an increment map and a non-verification indication mark;

any OSD component calculates a non-verified full-scale OSD map based on the increment map carried by the first increment synchronous message; and updating the full OSD map version.

7. The system of claim 6, wherein the method further comprises:

any OSD component receives a second increment synchronous message carrying the increment map, the Cyclic Redundancy Check (CRC) value and the check indication mark;

any OSD assembly calculates the total OSD map to be checked based on the increment map carried by the second increment synchronous message;

any OSD assembly calculates a CRC value to be checked according to the full OSD map to be checked;

the any OSD component comparing the CRC value of the second increment synchronous message with the calculated CRC value to be checked;

when any OSD component is determined to be inconsistent, the callback function process is not executed, and a full-quantity synchronization request message is sent to a monitor component which is used as the main equipment and sends the second incremental synchronization message;

the monitor component serving as the main equipment sends a full synchronization message with the full OSD map of the new version to the OSD component;

and any OSD assembly receives the full synchronization message, acquires the full OSD map of the new version and updates the full OSD map of the new version.

8. The system of claim 7, wherein the method further comprises:

the any OSD assembly determines that the CRC value of the second increment synchronous message is consistent with the calculated CRC value to be checked, and executes a callback function process;

and the OSD component is updated to the full OSD map of the new version.

9. The system of claim 6,

the monitor component serving as the main equipment acquires a cluster version of the distributed file system;

and the monitor component serving as the main equipment sends the first incremental synchronization message when determining that the version number of the cluster version does not correspond to the version number of the full OSD map of the new version.

10. The system of claim 9,

and the monitor component serving as the main equipment sends the second incremental synchronization message when determining that the version number of the newly acquired cluster version corresponds to the version number of the new version full OSD map.