Quiescing Update Activity
Whenever it becomes necessary to quiesce access to an area during a recovery operation, the quiesce must apply to all members of a data sharing group. For recovery purposes, the quiesce will usually be done by varying the area status to OFFLINE or TRANSIENT RETRIEVAL using a DCMT VARY command. This command must be executed in every member to establish a group-wide quiesce for a shared area. To do this easily, the command may be broadcast to other members of the group.
Occasionally, it will be sufficient to quiesce update access to an area through the use of a DCMT QUIESCE command. This command will automatically propagate the quiesce for a shared area to all group members, so there is no need to execute it on more than one member of the group.
Note: For more information about DCMT commands and how to broadcast them, see the CA IDMS System Tasks and Operator Commands Guide.
Recovery from a Warmstart Failure
If warmstart fails for one or more members of a data sharing group, recovery can proceed just as if the DC/UCF systems were not data sharing members. This is true even if manual recovery is necessary provided that all shared areas being updated by the members at the time of failure are quiesced in the remaining members of the group. The quiesce should be done by varying the status of the affected areas to OFFLINE using a DCMT VARY command.
If manual recovery is necessary, each member can be recovered independently provided that:
Failure to comply with these conditions, may result in database corruption.
Recovery from Other Types of Failures
Except when following the above procedure to recover from a warmstart failure, the archive files from all members that have updated a shared area since the backup was taken must be included in any manual recovery. Furthermore, the journal images from all members must be processed together in the same execution of the ROLLFORWARD or ROLLBACK utility. It is not valid to process the images from one member in one execution followed by the images from another member in another execution, since journal images must be processed in chronological sequence.
Recovery from a warmstart failure is an exception to this rule only because records updated by one member cannot be accessed by another member until the changes are committed or rolled out. If warmstart fails, the unrecovered records remain locked, so no other member can update them. This means that there will never be more than a single member with a before image for an unrecovered record and so inter-member sequencing is not important.
Using MERGE ARCHIVE
The MERGE ARCHIVE utility is used to merge the journal images from multiple members so that they are in chronological sequence. As noted above, most recovery utilities require that journal images be processed chronologically. In a data sharing environment, the journal images produced by each member are in chronological sequence, but the images for areas concurrently updated by multiple members are contained in each member's archive files. The MERGE ARCHIVE utility interleaves the journal images from multiple members so that they occur in date/time sequence. The resulting output file may then be used as input to a ROLLFORWARD, ROLLBACK, or EXTRACT JOURNAL utility statement.
When executing the MERGE ARCHIVE utility statement, the input consists of a concatenated set of archive files and optionally a merge archive file produced from a previous execution of the MERGE ARCHIVE utility. Archive files produced by a single member must be processed in the order in which they were created. Archive files from different members may be processed in any order relative to those of other members.
When to Use MERGE ARCHIVE
The output of the MERGE ARCHIVE utility can always be used as input to the ROLLFORWARD, ROLLBACK, and EXTRACT JOURNAL utility statements in place of the original archive files. It can also be used to combine local mode journal files and archive files when mixed-mode updates must be recovered.
However, while optional in most cases, MERGE ARCHIVE must be used to merge the journal images of multiple data sharing group members before those images are processed by:
Using EXTRACT JOURNAL
The EXTRACT JOURNAL utility is used to preprocess journal images in order to reduce recovery time. This utility can also be used in a data sharing environment. Any of the following are valid approaches to its use:
You must use the third approach if the ALL and STOP TIME parameters are specified on the EXTRACT JOURNAL utility statement; otherwise, any of the above approaches can be used to preprocess journal files in a data sharing environment.
If using either of the first two approaches, the EXTRACT JOURNAL utility statement can be executed on a periodic basis to preprocess the archive files created since its previous execution, or since a backup was taken. If recovery becomes necessary, all extract files produced since the backup must be concatenated as input to a single execution of the ROLLFORWARD utility. The order in which the extract files are concatenated must be such that the journal images for each member are in chronological sequence. It makes no difference in which order the images of one member occur in relation to those of another member.
If using the third approach, the entire set of archive files produced by group members that have updated the affected areas must be merged prior to executing the EXTRACT JOURNAL utility. The MERGE ARCHIVE utility can be executed on a periodic basis to merge the archive files created since its previous execution with the previously created merge file. The EXTRACT JOURNAL utility can then be used to preprocess the final merge file.
Note: For more information about executing both the MERGE ARCHIVE and the EXTRACT JOURNAL utility statements, see the CA IDMS Utilities Guide.
Coupling Facility Failures
Certain types of failures are unique to a data sharing environment, such as the loss of a coupling facility or a structure within the coupling facility. In some cases, all members of a group will fail and recovery must be coordinated across the group, a process called "group restart."
Note: For more information about recovering from coupling facility failures and group restart, see the CA IDMS System Operations Guide.
|
Copyright © 2014 CA.
All rights reserved.
|
|