Sie sind auf Seite 1von 6

Pinned Log Tail Recovery Log mode ROLLFORWARD Technote (FAQ) Problem With the TSM servers Recovery

Log in ROLLFORWARD mode, can the recovery log tail be pinned? Solution When the Recovery Log Mode is set to ROLLFORWARD, the "tail" of the log (the old est transaction record in the log) is always pinned in place by the last Databas e backup. If a Database backup is currently running, then the last transaction r ecord is pinned by the Database backup process. If a Database backup completes (scheduled, manual, or triggered by the dbbackupt rigger) and the utilization of the Recovery Log does not fall, and the completio n of the Database backup is accompanied by a message ANR4556W: (Excerpt From the Messages Manual) ANR4556W Attention: the database backup operation did not free sufficient recovery log sp ace to lower utilization below the database backup trigger. The recovery log siz e may need to be increased. Explanation: A database backup process has completed. The server is running in r oll-forward mode and the recovery log utilization remains above that specified i n the database backup trigger. System Action: A database backup operation will be triggered. Unless this situat ion is corrected, database backups will be triggered rapidly, one after the othe r. User Response: If a database backup operation does not free up enough recovery l og space to reset utilization below the database backup trigger, the recovery lo g needs to be extended or the database backup trigger updated to a higher log fu ll percentage. (End of Excerpt From the Messages Manual) In addition to the remedies in the Messages manual excerpt above, you may look f or a client session that has been active for a long period of time (session numb er will normally be much lower than other current sessions) or a server process that is active for a long period of time (again process number will be lower tha n current processes). Either of those cases may be holding an active transaction that disallows the Database backup to release the recovery log space. Cancellat ion of either session or process may free the record and allow the next triggere d database backup to clear the log. If it does not or there is insufficient time for a subsequent database backup to complete, then a HALT of the TSM server may avoid a crash of the server process. Be aware that when the server restarts a database backup will be started automat ically.

IC42909: ITSM WINDOWS CLIENTS BACKING UP SYSTEM OBJECTS CAN SUFFER PERFORMANCE P ROBLEMS THAT CAN EXTEND THE BACKUP WINDOW APAR status Closed as program error.

Error description Backing up Windows System Objects can lead to extended backup times, resource timeouts, pinned recovery logs due to some extra processing being performed on the ITSM Server. Some of the symptoms associated with this performance issue are as follows:

. - ITSM Server pinned recovery log imposed by Windows Clients - Extended Backup Windows - ITSM Server resource waiter aborts described in APAR IC36769 . This APAR only addresses potential performance issues during the backup of system objects and does not directly affect the processingof system objects via other ITSM Server processes such as expiration, migration, etc... . NOTE: This APAR does not account for all potential performance problems that a customer may encounter during the backup of system objects. This APAR only eliminates some of the extra processing that occurs at the end of the system object backup processing. . Platforms affected: All ITSM Servers at 5.1.X and 5.2.X . Customer/L2 Diagnostics . If a customer is suffering from any of the symptoms described above and they are backing up Windows system objects during that timeframe, it is recommended that the customer apply this APAR. . Initial Impact: Medium . Additional Keywords: txn commit vers versioning dirty page resourcetimeout sysfiles groups count endtxn anr0918E anr0538I Local fix 1 - Do not include system objects in the normal backup. They can be excluded by: Using the domain statement: DOMAIN ALL-LOCAL -SYSTEMOBJECTS Or Using the exclude statement: EXCLUDE.SYSTEMOBJECT SYSFILES . 2 - Backing up the system objects later using dsmc -optfile=xxxx where the optfile has a resourceutilization set to 1 so that the backup is single threaded . 3 - Limit the number of system objects being backed up during the same backup window. The likelyhood of this performance problem happening on the ITSM Server is increased with each Windows client added to the same backup window. Problem summary **************************************************************** * USERS AFFECTED: TSM servers for any platform supporting *

* the backup of SYSTEM OBJECTS from * * Windows clients or those TSM * * servers supporting a client that * * is using the TSM peer grouping support. * **************************************************************** * PROBLEM DESCRIPTION: See error description. * **************************************************************** * RECOMMENDATION: Apply fixing level when available. This * * problem is currently projected to be fixed * * in level 5.2.4. Note that this is subject * * to change at the discretion of IBM. * **************************************************************** The TSM server processing for peer groups, such as SYSTEM OBJECTS from Windows clients, has been modified to correct the inefficiency in the algorithm that can result in additional database I/O. Problem conclusion The TSM server processing for SYSTEM OBJECTS from Windows clients or other clients that perform peer group operations may incur additional database I/O which can impact performance. This additional database I/O is an inefficiency in the implementation of the current algorithm. This additional database I/O occurs during the client backup operation. Temporary fix Comments APAR information APAR number IC42909 Reported component name TSM SERVER 510 Reported component ID 5698ISMSV Reported release 52S Status CLOSED PER PE NoPE HIPER NoHIPER Submitted date 2004-09-29 Closed date 2004-11-22 Last modified date 2004-11-22 APAR is sysrouted FROM one or more of the following: APAR is sysrouted TO one or more of the following: PQ96732 Modules/Macros Fix information Applicable R51Z PSY R52A PSY R52H PSY R52L PSY R52S PSY R52W PSY component levels UP UP UP UP UP UP

IC34009: SHOW LOGPINNED

APAR status Closed as program error. Error description A new "undocumented" command is being added to the TSM server to help diagnose problems with a "pinned" recovery log. The TSM manuals will not be updated to reflect this command, but the command will be supported within the context of why this command was created and the limitations on this command. The new command is SHOW LOGPINNED In general there are five reasons why the recovery log is pinned. 1 - There is a transaction associated with a session or process that is using space in the recovery log and that transaction has been processing for some time, and while this transaction is processing space in the recovery log is being used buy this transaction and other transactions. Space used by other transactions that started after this transaction has started is still "used space" in the recovery log until this log running transaction completes even if those other transactions complete. 2 - The logmode is set to rollforward and space can not be freed until the next database backup. 3 - A database backup is currently running and there is data in the recovery log that the TSM server still needs as part of the database backup processing. 4 - There is a dirty page in the database buffer pool that needs to get flushed out to disk before the log tail can be advanced. 5 - The log tail can not advance until the next checkpoint. A checkpoint is something that the server handles automatically as space is used in the recovery log. Some terminology: What is a "pinned" recovery log? A pinned recovery log is when the "log tail" can not be moved because there is information needed in the recovery log for a transaction that is currently processing, for a transaction that has completed but has not been flushed to disk, for a currently running data base backup, or for a possible "roll-forward" operation on a data base restore. What is the "log tail"? The log tail is the point in the recovery log that contains the oldest record still needed by the server. What is the "log head"? The log head is the point that the next record will be written to the recovery log. When this command is issued the server will determine if one of the first four situations exists. If none of the first four situations exist then the server will report that the recovery log is not pinned and space will be freed with the next checkpoint operation.

If the second, third or fourth situations exist then the server will report which one of those conditions is causing the space to be pinned in the recovery log. If the first situation exists then the server will attempt to determine which session or process is associated with the transaction that is pinning space in the recovery log. In the process of determining this there will be some additional information reported that can be useful if the server is unable to determine a session or process assocation with that transaction. If the server is able to determine the session or transaction it will report the session or process number in addition to some other information such as the node name or the command issued for the process. This command is only intended to be run when the recovery log is actually pinned, therefore it is not recommended to schedule this command to run on a regular basis. If the recovery log is not actually pinned there are some small timing windows where incorrect results may get reported. Local fix Problem summary **************************************************************** * USERS AFFECTED: All TSM users having to diagnose a problem * * with a recovery log that fills too fast. * **************************************************************** * PROBLEM DESCRIPTION: There is currently no way to determine * * the exact cause for space in the * * recovery log not being released. * **************************************************************** * RECOMMENDATION: * **************************************************************** A new diagnostic command is being added to the TSM server. The command is SHOW LOGPINNED. See the APAR opening text for details on this command. Problem conclusion Apply this fix when available to have a new command SHOW LOGPINNED to help diagnose problems with a filling recovery log. Temporary fix Comments APAR information APAR number IC34009 Reported component name TSM SUN/SOL SER Reported component ID 5698TSMSU Reported release 420 Status CLOSED PER PE NoPE HIPER NoHIPER Submitted date 2002-06-28 Closed date 2002-06-28 Last modified date 2002-06-28 APAR is sysrouted FROM one or more of the following: IC33830 APAR is sysrouted TO one or more of the following: Modules/Macros

Fix information Fixed component name TSM SUN/SOL SER Fixed component ID 5698TSMSU Applicable component levels R420 PSY UP

Das könnte Ihnen auch gefallen