HP NonStop RDF, Руководство по управлению

Страница: 125 / 478

Exceeding the Maximum Number of Concurrent File Opens
The maximum number of audited files a single updater can have concurrently open is 3,000. If
you have more than 3,000 audit files being replicated by a single updater, then it is possible that
the updater associated with the volume may report RDF event 813 - "Concurrent file opens
exceeds capacity". This happens if the updater has 3,000 files open and it must open a new file.
Should this occur, the updater immediately generates the RDF event 813, commits its current
transaction, closes all files, and restarts, which generates RDF event 837. When it restarts, it
resumes processing image audit at the audit record for the file that caused the problem. In this
sense, the problem is self-correcting, it does not impact updater performance, and it is safe for
you to have more than 3,000 audited files on the volume. The danger comes when you have more
than 3,000 audited files on the volume and all of them need to be updated every 6-10 minutes
on a regular basis. If this happens regularly over a period of time, then it could cause the purger
to stop purging files.
When this situation occurs, you must stop RDF as soon as possible and rebalance the number of
audited files on the primary and backup volume of the affected updater so that you have no
more than 3,000 audit files on that volume. When RDF has been stopped and you have rebalanced
the audit files, then reinitialize and reconfigure RDF using the INITTIME option. See
“Initializing
RDF Without Stopping TMF (Using INITTIME Option)” (page 80) .
If the above problem occurs, the purger has stopped purging files, and you are unable to stop
RDF to rebalance the number of audited files on the volume, you can try lowering the duration
of the updater's transaction to the minimum value of 10 seconds as a short term workaround. If
this does not correct the problem, then the easiest way to correct the problem is to suspend the
extractor on the primary system for 10 minutes. If you have RDF/ZLT protection, then you are
not at risk of losing any data if your primary system should fail while the extractor is suspended.
If you do not have RDF/ZLT protection, then you are vulnerable to loss of data to an unplanned
outage of your primary system from the point where you suspend the extractor to the point
where the extractor has caught up after you have activated it, but this workaround will allow
the purger to purge files.
If suspending the extractor is not acceptable, then use the following steps to resolve the problem
for the short term:
1. Do status RDF to get the name of the latest image file on the image trail of the updater
generating the 813 events.
2. Stop RDF
3. Restart RDF with UPDATE OFF; this causes the receiver to rollover to a new image file on
each image trail.
4. After one minute, STOP RDF again
5. Restart RDF with UPDATE OFF; this causes the receiver to rollover to a new image file on
each image trail.
6. On the image trail for the updater generating the 813 events, move the next file in sequence
(the one after the file identified in step 1) to a different subvolume. For example, if the updater
is reading file AA000100, then move AA000111.
7. Move the file moved in the previous step back to the image trail.
Remember that all of this can be avoided by keeping your audit files balanced so that you do
not have more than 3,000 on any single RDF-protected volume.
Responding to Operational Failures
RDF can recover from any of the following events, as described in detail in the following pages:
• Communications line failure on the primary or backup system
• System failure that does not require an RDF Takeover operation
• Processor failure on the primary or backup system
Exceeding the Maximum Number of Concurrent File Opens 125