WO2010148988A1

WO2010148988A1 - Method, device and system for taking over fault metadata server

Info

Publication number: WO2010148988A1
Application number: PCT/CN2010/074042
Authority: WO
Inventors: 程菊生; 徐涛; 陈浩; 钟吉林
Original assignee: 成都市华为赛门铁克科技有限公司
Priority date: 2009-06-24
Filing date: 2010-06-18
Publication date: 2010-12-29
Also published as: CN101577735A; CN101577735B

Abstract

A method, device and system for taking over a fault metadata server is provided in the embodiments of the present invention, wherein the method is applied to a metadata server cluster, and the metadata sever comprises a local metadata tree and a neighbor metadata tree; the local metadata tree manages the local file system, and the neighbor metadata tree corresponds to the local metadata tree of the neighbor metadata server. Said method includes the following steps: when the neighbor metadata server corresponding to the local metadata server is in working order, the local metadata server performs real-time mirror image on the local metadata tree of the neighbor metadata server by the neighbor metadata tree; when a fault occurs in the neighbor metadata server, the local metadata server, by managing the neighbor metadata tree which has been performed real-time mirror image on, takes over the neighbor metadata server with fault. Improving the responsibility of the metadata server is achieved by the technical solution in the embodiments of the present invention, further improving the responsibility of the whole distributed file storage system.

Description

The method, device and system requirements for taking over the fault are submitted to the China Finance Bureau in 2009 6 24

200910150732.8, "The method, device and system for taking over the faults", the first of its contents, the contents of which are incorporated in this section. Wood field

This paper deals with the field of distributed file storage, especially the methods, devices and systems for taking over faults (eadaaSe Ve , ). Background

Nearly undistributed files have been heavily developed for storage solutions. Since the number of files managed by the distributed file system is very large, there are upper-level files or dozens of files that need to be managed by each line. Look at the storage mode of the large and storage (storage refers to (se), wood or distributed file system and other functions, the collection of a large number of different types of storage software to work together, provide storage and functions of the common system) The sheep are no longer able to meet the needs of users, and need to use multiple components of unmanaged

.

In the existing woods of the present invention, at least the following distribution files are distributed to each customer, each storage, and each storage (ec Soage SeVe, OS). All management systems are owned. If a failure occurs, the distribution file will not work. Looking at the increase of the distribution file system, the more the number of each is, the less the probability of each failure is, and the reliability of the system is obtained. Therefore, how to improve the reliability of the distributed file storage. Content

This provides methods, devices, and systems for taking over faults to improve reliability and improve the reliability of distributed file storage systems.

Aspect, this method provides a method for taking over the fault, Each (SC se ), including local and local local files, is managed locally.

The method includes the local working of each device in the local area, and the local image in each local device, and each local fault in each device is faulty, local

Like line management, to take over each of the faults.

On the other hand, this section provides means for taking over the faults, each of which includes local,

Locally managed locally, local devices include 奕吋羊羊 , , , , , , , , , , , , , ,

The local image in each device, with each local file in each device, the faulty takeover of the sheep,

Each fault occurs and the local image is managed to take over the faulty ones.

On the other hand, this provides the system for taking over the faults. The upper system includes the customers. Each of the elements in each unit includes local and local local file management, and each local and local. Everything works often, through

The local image in each device is managed by each local file in each device, and each local image is managed to take over the failed storage. Each of the media stores the elements in each device.

The wood plan has the following beneficial effects

In the operation of each local device, the fault occurs on each image, and the local image management is performed to take over the fault.

The various wooden means, so as to improve the reliability of each, and improve the reliability of the distributed file storage system.

Solid description

To explain more clearly the wood scheme in this book, the following is the introduction of the sheep that need to be used, and the following is the ones in the field, in the ordinary wood in the field, without paying for the work,

1 force this take over the fault of each method flow

2 Liben Forest Document

3 Forces

5 Liben's work under various faults and take-overs 6 Forces take over the faults of each device

7 Liben take over the fault

8 force example cluster case

9 Forces in the case of clusters, each fault and takeover situation

Hugh way

The following will be combined with the wood scheme in this book. , complete, of course, this part, not all. In this book, all the other common woods in this field are not guaranteed to be made.

As shown in Figure 1, the method of taking over the faults is in each, including local,

Local to local file management

Locally

The method includes the following steps

S 1, the local devices work normally, this image, each local s102 in each device, each failure, local

Like the management of the line to take over the faulty

Distribution of and , as well as local and

Mutual.

Every local device works often, sexual "like", local sexual force " "

"Image" and " " of each failure. Yes, each failure has occurred for a long time,

Force local, faulty local file management methods including, second, second

In the second local, in the second

Each of the lines is the second of each of the original locals. Various faults, local devices and villages Everything is normal.

The wood scheme of the above method has the following beneficial effects. The local devices are often used to pass through the meta-images of the upper devices in the local devices.

Each of the failures and the management of the local image are taken over to take over the various wooden means of the failure, thus improving the reliability. Take over the faults and improve the reliability of each

"Forest File System" is not managed. The forest file system is a general document system, 2, including

Forest frame ( a ewo ) and ( ead aTee )

The forest frame all the distributions in the system, as well as the mutual (sub, brother, etc.). Same as each

Assign a local file system. The forest framework can be managed much.

It is a forest file system consisting of local files. The tree is managed by each. The operations are not completed.

Operation can be below

(1) " " (Read- e) can be operated

(2) Only "(ead-O) only, can't and other operations

(3) "Image" (o-O) can only be operated like the image of . This image refers to the number of copies.

The location can be different (oca ead aTee, local) and (eghbo ead aTee, ).

The local refers to the yuan that is stored locally.

Means that each number is saved in the local area, the same image

Each of them, that is, the element stored in S, the other element. Local failure occurred and was not taken over. Often, only have "like" , no ". take over, special cost-changing, with local all.

2, Liben Forest file is generally indicated, in 2, including 3 yuan, respectively 1, 2 and 3. Each of the number of media is unmanaged, and each S-, S-2, and S-3. For the rest and the faults, you can put the composition, "( eeR )", that is, in each of the above, the local, the local, the phase, and the 相, do not appear local or more, Therefore, the same can be seen as a mediation. 2, the elements shown in the local, the next 1

1 local,

Local

S-1 1 3

S-2 2 1

S-3 3 2 is 1, local is 1 in each S-, 3 is the same, 3 is also local in S-3, that is, S-1 is local in S-3, and S- 3 If a fault occurs, the S-1 (3) line manages the takeover of the S-3 that has not failed.

It should be noted that the scheme of the above scheme can also be managed by multiple (2, or 3), and the management method can refer to the phase steps in this section.

3, this takeover failure process steps under S301, forest framework

The forest framework has all the distributions in the system, and all The mutual (sub, brother, etc.), which is based on the local file system, has the management and operation of the file, mercury and sub-responsibility. Same, you can separate each of the four in the tree, the next 2, the force points

2

Each

S

S1 small o e

S2 / s

S3 /ec

S4 / v in 2, the division of each S, that is, the road to S to "/ho e", the rest of the appliances can be 2. It should be noted that the mercury can be used.

S302,

4, the outline of the forest document of the forest,

(1) Local

Figure 4 shows the 5 of the forest framework of the Liben Forest Document System. It is the tree of the forest file system, and this is the local file system. The tree is managed by each. The operations of , , and so on are unfinished. Is a local file system, by local

( oo od ), index ( od ), ( aaboc ). No, the local file system is a forest file system.

0 (2) of

It is the local image of each other. In all normal circumstances, only have "images", only use images, and no "" .

For each local, you can refer to 1 in the section.

Therefore, if the local area is too large, the forest frame can be re-planted on the new yuan, and the local part of the mercury is moved to the village to form a new yuan, local, same, and other.

5303, the local devices work normally, the general case, the local system shown in 2, the S-3 in the S-

For example, the local image in S-1 in S-2 and the local image in S-2 in S-3. Like, can

Each local file is governed by each.

5304, each failure, unmanaged

Each of the elements can be the same, the case 2, in S-, each refers to the yuan, that is, each 3 (3).

Each failure, for example, each 3 (S-3) fails, will be "like" "", this, S-1 can

3 operations, not 3 yuan Each line 3 management. Yes, can include steps

S305, each long-term, special local, generate new second line management

For example, 2 and 5, S-1 fails and does not exist for a long time, like S-1's local S-2 force local -2, take over S-1

management.

Same, new, not the second line management, . For example, S-2 generates a new local image of S-3 (S-3 original S-), and S-3 generates a new -2 S-2 local-2 image.

Yes, can include steps

5306, each, local and

If it is constant, then each

Everything is constant. For example, the 5 shown in Figure 5 will be shown.

This paper proposes a method in which Shentong does not take over the faults, and each local device works normally.

Each of the local images of each of them fails, and each local takes over. This paper proposes a method of "forest file system local". The forest documents are unified, local, local, and so on. The beneficial effects of the wood scheme greatly improve the reliability of the distributed file system. This will greatly improve the reliability of the distributed file system. If each of the faults occurs (as long as the faults are the same), the system can not take over the faulty, and the availability is unresolved. As long as the same fault does not occur (in the same time, the probability of occurrence of each fault is also small), in the case (most often, all other

Each file is faulty, and the distribution file is still available. The method of giving God has greatly improved

Reliability. As shown in Figure 6, the devices that take over the faults are shown in each, including local and local local file management. Included in each local device 70

像 Like the sheep 701, the local devices work normally, and the locals in each device

For example, each faulty receiver 702 is managed by a local file in each device, and each faulty, local image is managed to take over the faulty ones.

Yes, the device 70 includes a forest frame-forming sheep, a distribution of the forest frame, and a local and mutual image.

Every local device works very often.

Sexuality "like", local sexuality"

Each of the failures will be "like" "". Yes, the faulty takeover of the sheep 702 in each of the faults, long-term, local, faulty

Each local file management fails to take over the sheep 702 in the second, second in the second local, in the second image

The second of each local original

.

Device 70 includes faulty sheep, local devices and .

, which improves the reliability of distributed file storage systems.

inverted

7 shows the overall composition of the force takeover failure The upper system includes the client 81, and each of the elements in the 82 and the storage 83 stores each storage.

More than each of the 82, each of which includes local and local local file management, and the local local devices work normally.

a local image in each device to

The local files in each device are managed by each row of faults and local images, in order to take over the faulty memory stores 83, in each of them, to store the elements in each device.

The following is a combination of the image of Hugh’s image on the top.

1) Normal situation

As shown in Fig. 8, it is often indicated in the case of the cluster. This distribution and storage system is divided into three major customers, each, and each storage. Each consists of multiple components. Each of the responsibilities is stored separately.

2) Fault and takeover

Figure 9 shows the symptoms of each failure and takeover in the case of the cluster. In each of the faults, the method proposed by the system is replaced by a failure. Since all the latest elements and mercury have been saved on the faults, it is only necessary to take over the faults, and only the """ can take over the faults.

3) Troubleshooting and number

10 shows the obstacles in the case of the cluster in the case of the case and the number of cases. Each fault has been eliminated, and the proposed method will be re-organized to bring the latest new city to the local. The new city has not established a local, and the locals are all local and local.

The wood scheme of the present system has improved the reliability of each and improved the reliability of the distributed file storage system. Yes, the program is not commanded. The hardware is not completed. The program can be stored in the storable program, including all or part of the steps, storage, RO / , and so on.

The above-mentioned rest mode, the purpose, the wood plan and the beneficial effects are explained in detail, and it is understood that the above-mentioned method of rest is not used to limit the warranty, any modification made in the spirit and principle of the present. , equivalent replacement, etc., are included in this warranty.

Claims

Requirements 1. Each method of taking over a fault, which is characterized in that each local, local, local, local file is managed locally, including

Each local device works normally, local image, and each file in each device

Each failure, local

Like the line management, to take over the faulty ones.

2. Requirement 1 method, which is characterized by the distribution of the distribution, as well as the local and mutual.

3, requires 2 methods, which are characterized by

The local devices work normally.

Sexuality "like", local sexuality"

"Image" and " " of each failure.

4. The method 3 is required, and the method includes

Occupy each failure, long-term

Force local, faulty local file management

Second, second, second

Each local, in the second image

The second of the locals originally had their own.

5, request 1 method, its special Each fault, the local device and the.

6. Take over the faulty devices, which are characterized in that they include local and local local file management, including in each local device.

像 Like a sheep, working in each local device, usually in the local

Like, using local files in each device

Failure to take over the sheep, in each failure, local image management, to take over the failure

Each.

7. Requires 6 devices, the special features of which include

Forest frame generation sheep, forest frame generation, forest frame preservation

The distribution of local and local, as well as the local and mutual.

8, requires 7 devices, which are characterized by, like sheep

Every local device works very often.

Sexuality "like", local sexuality"

Each of the failures will be "like" "".

9. Requires 8 devices, which are characterized by

Failure to take over the sheep in each failure, long-term, local, failure

Local file management

Faulty take over the sheep in the second, second

In the second local, in the second

Each of the lines is the second of each of the original locals.

10. Requires 6 devices, the special features of which include

The faulty sheep are common in the local units and in the village.

11. Take over the faulty system. The special features include the customer, each of the yuan and each of the yuan, each of which includes local and local local file management, local and local. Everything works often, through

The local image in each device is managed by each local file in each device, and each local image is managed to take over the obstacles.

Each of the storage elements is stored in each of the cells.