High CPU Utilization from searches
We have started having issues with the poa spiking the CPU for up to 8 hours. I found all GWTCP handler threads being allocated then "hanging" for up to 8000 seconds. I looked at the http monitor to determine what the threads were doing. The same user had all the threads and they were all doing the same thing as indicated below by a snipit from the thread activity:
Current Transaction Destroy streamed search
Time Elapsed (Secs) 6483
The user had been logged into webacc and did some searches. His mailbox has over 100,000 messages (I know...archive but mgmt. hasn't listened). When I spoke to him he said that the searches did not complete and he logged out.
I feel confident that this has something to do with QF indexing not completely indexing his email. The qflevel was at default and the POA log showed about 90k of unindexed email. I switched the qflevel to 999 and restarted the POA. It has been running at 999 for two nights now. He has not logged into webacc and done searches since so I am not sure if this fixed it.
The user's db seems fine ( I run a structure check nightly and there are no errors). I did a validate on both the domain and post office databases and they were fine.
My questions are:
1. POA logging is set to normal and not all users are shown in the logs during the QF indexing period. Is this normal or does the poa need to be logging at verbose to show all users?
2. Does this seem like the correct approach to this issue?
Thanks,
Mike
|
Threaded Mode
