JPH0628042B2 - How to update the journal version - Google Patents

Description

DETAILED DESCRIPTION [Outline] When a predetermined event occurs, an equivalence guarantee process for equalizing data in a plurality of physical volumes forming a logical volume is performed, and an equivalence guarantee process is performed. The system calculates the base version number based on the latest version number of the journal written in each physical volume, and notifies the other system in operation of the calculated base version number.

[Industrial application field]

According to the present invention, when equivalence guarantee processing for equalizing all data of a plurality of physical volumes constituting a logical volume is performed, the system in which the equivalence guarantee processing is performed is written in each physical volume. The present invention relates to a method for updating a journal version number in a shared multiplex volume system in which a base version number is calculated based on the latest version number of and the calculated base version number is notified to all other operating systems. .

[Conventional technology and problems]

FIG. 7 is a diagram for explaining the multiplexed volume. In FIG. 7, PV1 and PV2 are physical volumes, LV is a logical volume, R is a record, and T is a track. Logical volume LV is physical volume PV
1 and PV2, the physical volume PV
The same data is stored in 1 and PV2. The physical volume PV1 is divided into a user area and a system management area. The user area stores data used by a user program, and the system management area stores data used by an operating system such as a journal. The physical volume PV2 has the same configuration as PV1.

In order to make the two physical volumes PV1 and PV2 appear as one volume (logical volume), the data equivalence of the physical volumes PV1 and PV2 must always be guaranteed. As a method of guaranteeing the equivalence when the logical volume LV is first accessed, there is equivalence assurance processing by a journal. This will be explained below. As shown in FIG. 7 (b), the logical volume LV
Is partitioned into partitions and when the partition is updated, the journal shows that the partition is being updated. Although the logical volume LV is divided into 16 in FIG. 7B, this is merely an example, and the actual number of divisions is larger than this. Journal is physical volume PV
Since it is created in the system management area of PV1 and PV2, even if the system is stopped and the data in the main memory is lost, the updated part (the part where data equivalence cannot be guaranteed by reading the journal on the medium ) Can be known, and data equivalence can be guaranteed by copying from the reference physical volume to another physical volume.

The version number is added to the journal to show its freshness, and the version number of the journal is incremented each time the journal is updated. This will make the journal with the highest version number the latest journal. Searching for the latest journal requires reading the journals of all physical volumes and finding the one with the highest version number.

Writing to the journal is necessary every time the data is updated, and reading of the journal is required to obtain the latest journal. Therefore, although it is usually necessary to write a journal when updating data, the overhead of journal writing becomes an important problem in the performance of the multiplexed volume. There are the following software solutions for reducing the time required to write to the journal. The biggest problem with I / O to the journal is the process of looking for journal records. To minimize and average this, as shown in FIG. 7 (c),
It suffices to create a journal record R on a predetermined journal writing track T without any gap and write the journal to the nearest journal record R. By doing this, there is almost no waiting for rotation. When reading, you can read all journals, find the one with the highest version number, and consider this as the latest journal written on that physical volume.

The above is a description of a conventional multiplexed volume journal. Based on this, a problem when expanding to a shared multiplex volume will be described. FIG. 8 is a diagram showing an outline of the shared multiplex volume. In FIG. 8, 1 and 2 are systems, and A and B are intersystem communication devices. Each of the systems 1 and 2 is composed of a central processing unit, an operating system, etc., and uses the logical volume LV to perform business. The intersystem communication device A is for exchanging multiplexed volume control information, and the intersystem communication device B is for system operation / stop monitoring.
For safety, the intersystem communication devices A and B use different communication means, for example, the intersystem communication device A uses FSI and the intersystem communication device B uses shared DASD. Note that FSI is a type of communication method using an optical communication cable. When sharing a multiplexed volume (synonymous with a logical volume) and sharing a journal between systems, the following points become the overhead of the input / output processing for the journal.

(a) Journal writing and exclusive control of I / O processing that causes it (b) Intersystem communication for changing the journal version As a method of solving these, for example, as a journal writing track for the system 1. T1 is assigned, and T is assigned to the system 2 as a journal writing track.
There is a method of allocating 2 and allocating T3 to the system 3 as a journal writing track. in this way,
If each system has a journal writing track, the latest version will be unique to each system, and the inter-system communication of (b) will not be necessary. With regard to the above-mentioned (a), it is sufficient to do the same as the input / output with respect to the normal shared volume, and there is no particular overhead as a shared multiplex volume.

Next, consider reading a journal. It is necessary to read the journal in the following two cases.

i) When it is necessary to know the partition being updated on the logical volume in order to guarantee equivalence ii) When the version number of the journal to be written is determined in order to start using the multiplexed volume. Read the journal. The process is very time consuming and
Since a large buffer is required, the system performance is deteriorated by occupying the memory. The reading of the journal is a process necessary to guarantee the equivalence at the time of initializing the multiplexed volume, but it is useless for each system to perform it individually.

[Object of the Invention]

The present invention is based on the above consideration, and in a complex system in which a multiplex volume is shared by a plurality of systems, shared multiplex volume that enables to omit journal reading at the start of use of the shared multiplex volume as much as possible. It is an object of the present invention to provide a method for updating the journal version in a computerized volume system.

[Means for achieving the purpose]

Therefore, the journal version updating method of the present invention comprises a plurality of physical volumes each having a user area and a system management area and is divided into a plurality of logical volumes, and a plurality of logical volumes sharing the above logical volumes. The system management area has a track for journal writing for each system, each system updates the current journal version number when writing a logical volume, and Write the journal information having the display during update, the update partition indication information and the journal version number to the closest one of the plurality of journal records existing on the journal writing track of the physical volume user area. Data is updated, and after the data has been written, the update is displayed. A method of updating the journal version in a shared multiplex volume system configured to write a deleted journal, which is stored in the physical volume when the logical volume is used for the first time. Read all journals of all systems, check the partition being updated, perform equivalence guarantee processing to make the data identical for the partition being updated, and after performing equivalence guarantee processing, the latest journal version of all systems Of the maximum value of 1 is added to the base version number to notify all other operating systems, and the base version number is used as the initial value of the journal version number of the own system. When you try to use the above logical volume newly under the condition that it is being used in the system, the base version of other logical system in use is changed. Perform combined, the base version number sent from the other system is characterized in that the journal version number initial value for the self-system.

Example of Invention

In summary, the present invention is as follows. The first system to start using the multiplexed volume in the whole system reads the journal to assure equivalence and to determine the version number of the journal to be written. At this time, it is necessary to read the journals of the entire system in order to know the area being updated. Since the latest journal of each system has been found at this point, the system that uses the multiplexed volume can be notified later, and the process of determining the journal version number can be omitted except for the first system. Further, the equivalence guarantee process may be performed only when the equivalence may be broken, and a system that uses the multiplexed volume later does not need to perform the equivalence guarantee process.

If the version number at the start of writing the journal is individually stored in each system, communication volume will increase due to the need to monitor all systems for a certain system (in consideration of startup after down), and the protocol will be complicated. Therefore, the initial value common to the system that newly uses the multiplexed volume is stored, and this is managed as common information for all systems. This initial value is called the journal base version number. The base version number is obtained by adding 1 to the maximum value among the maximum version numbers of each system. After that, when the system that is using the logical volume is stopped (down, etc.), the base version number is the maximum version number of the stopped system plus one. This eliminates the need to distinguish between the base version number notified to the system that newly uses the multiplexed volume and the version number notified to the system that restarts the use of the multiplexed volume (the system that started up after being down). Further, since the time when the base version number is updated and the time when the equivalence guarantee processing is performed coincide with each other, it is not necessary to read the journal before the time of guaranteeing the equivalence.

The relationship between the journal version and base version of each system is explained. The journal version number of each system is unique, and if the number of writes is different, it will be different over time even if it starts from the base version number. Here, the meaning of introducing the base version number is to omit the jar read of the system that starts the use of the logical volume and the system that restarts the use by the IPL after the down. Therefore, to ensure that the system that queries the base version for use will get the correct information when querying any system,
Notify all systems. Actually, the common management information includes not only the base version number but also the volume reservation information, so when the normal access to the logical volume becomes possible, the management information is shared with other management information. hand over.

Hereinafter, the present invention will be described with reference to the drawings. Fig. 1 (a)
FIG. 6 is a diagram showing processing of an input / output supervisor when a user issues an input / output request to a logical volume.

Check the request type. In the case of READ, the process of is performed, and in the case of WRITE, the process of is performed.

A CCW (channel control word) for displaying the writing in the journal is prepended to the CCW of the user. At this time, the version number of the journal is incremented by one. The current journal version number is stored in the main memory. The initial value is the base version.

Issues SIO (input / output device start-up command). Request is WR
In case of ITE, issue for all physical volumes.

FIG. 1 (b) is a diagram showing a CCW sequence issued at the time of WRITE. CCW for writing the latest journal is the user's CC
Beyond W, as a result, the data is updated after the journal is updated, and there is no omission in the equivalence guarantee processing based on the journal. After the input / output processing is completed, the journal whose display is being written is erased is written.

FIG. 2 is a diagram showing the equivalence guarantee process when the logical volume is first used in the entire system. The method of recognizing the first use of the logical volume in the entire system is performed as follows. The IPLed system inquires the management information of all the shared logical volumes at that time through intersystem communication. At this time, if it is unused, it is responded as unused. Therefore, it can be determined that the system is not used until its own system starts to be used or until another system notifies the start of use. The equivalence guarantee process when the logical volume is first used in all the systems will be described below.

Read the journal and search for the latest version.

Check to see if you have done it for all systems. If not done for all systems, go back to
If this is done for all systems, perform the process.

Logically sum the locations (partitions) that are being updated in all systems.
For example, when the logical volume is divided into 16 partitions, the update partition instruction information has a 16-bit configuration, and if partition 1 is being updated, the first bit in the update partition instruction information is the logical bit. It is set to "1".

Guarantee equivalence based on logical disjunction of journals. The equivalence guarantee is made by copying the part being updated.

Find the base version of the journal. The base version number is obtained by adding 1 to the maximum value among the latest version numbers of all systems.

Notify all operating systems of the obtained base version.

The version number of the journal to be written next is the base version number.

Accept requests from users.

FIG. 3 is a diagram showing the equivalence guarantee processing performed by the system that first uses the logical volume after that time when the other system goes down. In addition, an inter-system communication means is provided between the systems, and the data indicating that oneself is alive is transmitted to the other system at regular time intervals, and when this data is not sent from the other system, the other system is transmitted. Have stopped or went down.

Read the journal of the downed system. At this time, the input / output request from the user is temporarily held.

Equivalence guarantee processing is performed based on the latest journal of the downed system.

Check if the latest journal version is greater than the base version. If the version number of the latest journal is larger than the base version number, the process is performed. If the version number is not large, the process is performed.

The version number of the latest journal of the downed system plus 1 is used as the base version number, and the other system in operation is notified.

Receive user input / output requests.

FIG. 4 is a diagram showing processing performed by another system while another system is down and another system is executing the equivalence guarantee processing. It should be noted that the fact that the other system is down and that the equivalence guarantee process is being executed is notified to the other systems by the system that performs the equivalence guarantee process.

Hold user input / output requests.

Wait for completion of equivalence guarantee processing in other systems. When the equivalence guarantee process is completed, the base version number is notified.

Update the base version of the journal. Note that the journal version may be a continuation of the previous version.

Receive user input / output requests.

FIG. 5 is a diagram showing processing performed by a system that starts using a logical volume later when the logical volume is already being used by any system.

By the inter-system communication means to the system already in operation,
Queries the base version.

The initial value of the journal version used when writing is set to the notified base version.

Receive user input / output requests.

FIG. 6 is a diagram for explaining the timing of equivalence guarantee and journal reading. Initially, it is assumed that system 1 and system 2 are in operation. In the illustrated example, the system 1 is the system that first starts using the logical volume. The system 1 reads the journals of all the systems, performs the equivalence guarantee process, then finds the base version number, and uses this as the initial value of the journal version number, and notifies the system 2 of the completion of the equivalence guarantee process and the base version number. Then, the acceptance of the input / output request is started. The system 2 sets the received base version number as the initial value of the journal version number and starts accepting the input / output request of the user.
After performing the IPL, the system 3 makes an inquiry about the base version number to another system (system 2 in the illustrated example), and the system 2 notifies the system version number to the system 3. In this state, when the system 1 goes down, another system (system 2 in the illustrated example) detects the system 1 going down, prohibits the user's access, and notifies the system 3 of the start of the equivalence guarantee processing. To do. When notified of the start of the equivalence guarantee processing, the system 3 prohibits access by the user. On the other hand, the system 1 performs the IPL and inquires of the system 2 about the base version number. Since the system 2 has not yet completed the equivalence guarantee processing, the system 2 is notified of access prohibition. The system 2 reads the journal of the system 1, performs the equivalence guarantee process, determines the base version number, notifies the system version numbers to the systems 1 and 3, and resumes the request acceptance by the user. System 1 notified of base version
Sets the base version number as the initial value of the journal version number and starts accepting the user request. Also in the system 3, when the notification of the base version number is received, the acceptance of the user request is restarted.
It should be noted that in the systems 2 and 3, the journal version number may be a continuation of the previous version and does not need to be changed. In addition,
The systems 1 and 2 also inquire management information after IPL.
The management information referred to here is configuration information such as physical volumes incorporated in a logical volume, equivalence information such as base version,
Includes shared control information such as whether or not there is a reserve. Therefore, the systems 1 and 2 also make an inquiry. However, when IPLing for the first time through each system (which can be known by intersystem communication), the management information can be initialized by simulating or omitting the inquiry.

〔The invention's effect〕

As is apparent from the above description, according to the present invention, a system that starts using a multiplexed volume that is already used by another system does not need to read a journal. It is possible to prevent speeding up and system performance degradation. As mentioned above,
The process of reading the journal is very time consuming, and
Since a large buffer is required, the system performance is deteriorated by occupying the memory. In the conventional technology, the journals are read by each of a plurality of systems that use the shared multiplexing volume, and therefore the performance degradation is very large.

[Brief description of drawings]

FIG. 1 shows the processing of the I / O supervisor when the user issues an I / O request to the logical volume and the CCW sequence issued at the time of WRITE, and FIG. 2 uses the logical volume first in the entire system. FIG. 3 is a diagram showing the equivalence guarantee process at this time, FIG. 3 is a diagram showing the equivalence guarantee process performed by the system that uses the logical volume first after that time when the other system goes down, and FIG. Figure 5 shows the processing performed by other systems while the system is down and another system is executing the equivalency assurance processing. Figure 5 shows the logical volume after the logical volume is already being used by any system. Showing the processing performed by the system that starts the use of data, FIG. 6 is a diagram explaining the timing of equivalence guarantee and journal reading, FIG. 7 is a diagram explaining a multiplexed volume, and FIG. 8 is shared Is a diagram showing an overview of duplex volume. 1 and 2 ... system, A and B ... inter-system communication device, PV
1 and PV2 ... physical volume, LV ... logical volume,
R ... record, T ... track.

Claims (1)

[Claims] 1. A logical volume comprising a plurality of physical volumes each having a user area and a system management area and divided into a plurality of partitions, and a plurality of systems sharing the logical volume. The system management area has a track for journal writing for each system, and each system updates the current journal version when writing data to a logical volume, and for all physical volumes, it uses its own journal writing. Writes to the closest of the multiple journal records present on the track the journal information having the update indication, the update partition indication information and the journal version, and then the data in the user area of the physical volume. After updating and after writing the data, the process to write the journal whose display is being updated A method of updating journal version in a shared multiplexed volume system configured to perform the following: when first using a logical volume, read all journals of all systems stored in the physical volume, Check the partition being updated, perform equivalence guarantee processing to make the data identical for the partition being updated, and after performing equivalence guarantee processing, set the maximum value of the latest journal version of all systems to 1 Is added to the base version number to notify all other operating systems, and the base version number is used as the initial value of the journal version of the local system, and the logical volume is already in use by another system. If you want to use the above logical volume newly under The journal version update method, which uses the sent base version as the initial value of the journal version of the own system.