TMF Operations and Recovery Guide (G06.26+)

Occasional Operations
HP NonStop TMF Operations and Recovery Guide—522417-003
3-6
Responding to Audit-Trail Overflow
Reason: Long-running transaction
Meaning. All audit-trail files affected by a transaction must remain on the active or
overflow volumes until the transaction is committed or aborted. Normally, TMF
automatically aborts transactions that run longer than two hours; however, if the
autoabort function has been turned off, a transaction can run an excessively long time,
causing the audit trail to fill.
Action. Issue a STATUS TRANSACTIONS, DETAIL command to identify any long-
running transaction and determine what process started this transaction and when.
Determine whether it is appropriate for this transaction to be long-running.
•
If the transaction really should be long-running, then let the transaction continue to
run as long as possible, carefully watching the value for “Active audit-trail capacity
used” on the STATUS AUDITTRAIL display. If the audit trail is approaching the
begin-transaction-disable threshold, or if the transaction is not about to be aborted
anyway because of the 45% MAT capacity limit noted above, use the ABORT
TRANSACTION command to stop the transaction; this command also backs out
any effects the transaction had up to that point. Once the transaction is stopped,
the TMP releases the audit-trail files that were needed by the transaction, which
frees space on the audit trail. See The ALTER AUDITTRAIL Command on
page 3-7 for more information.
•
If the transaction should not be long-running, use the ABORT TRANSACTION
command to abort the transaction, and report the incident to your system manager.
If it is crucial for the transaction to complete, you can temporarily increase the size of
the active-audit trail. See The ALTER AUDITTRAIL Command on page 3-7 for more
information.
Reason: Data volume recovery mode configured as ONLINE
Meaning. When the recovery mode for a data volume is configured as ONLINE, the
audit-trail files that would be needed to recover that data volume must remain pinned
on the active-audit volume. If these files cannot be released to make room for new
audit records, the audit trail eventually fills.
Action. First make sure that both the active-audit volume and the data volumes are up.
If TMF operations are running normally and audit information is being generated at a
typical rate, consider changing the data volume configuration to use the archive
recovery mode: this way, audit-trail files do not have to remain on the active-audit
volume solely for the purpose of volume recovery (archive mode is allowed only if the
audit trail for the data volume is configured for audit dumping). See Specifying
Caution. TMF aborts any transaction pinning the oldest MAT file if the file is pinned because
of currently active transactions and if audit information is filling 45% or more of the MAT
capacity. Beyond this, however, when the autoabort feature is off (threshold set
to 0), there is no mechanism to abort runaway transactions automatically; if multiple runaway
transactions go unnoticed, they can cumulatively pin many audit-trail files, preventing them
from being recycled and prohibiting the start of new transactions.