Re: Oninit processes running away with process time

In article <3419786A.39F5@bloomberg.com>, "Art S. Kagel" <kagel@bloomberg.com> writes >Robertson Lee NTSE wrote: >> >> Hi, >> >> We are running Informix 7.13 on HPUX 10.01 and we are currently >> experiencing a problem with 2 of our oninit processes. At the OS level >> they are eating up a large amount of CPU time. Does anyone have any >> ideas as to what this may be or how we can find out what they are doing. > >Do you have two listener threads configured? If so this may just be >polling overhead. You can check by running onstat -g ath for threads >whose name (last column) includes "poll" the vp-class column tells you >which VP contains that thread. Onstat -g glo will tell you the pid of >the individual VPs. If you have two poll threads probably their pids >are the ones getting so much CPU time. This is normal. > >Art S. Kagel Try onstat -u first. DO any users have a large percentage of reads? E.g. most users<1,000 reads, 1-3 users at >10,000. If so these users are doing selects without using an index. Run select hex(partnum),pf_seqscans from sysmaster:sysptntab and compare against select hex(partnum),tabname from systables order by 1 ^^^^^^^^^^^^ take the last few digits from here and compare with the first select to get the name of the table being scanned. Also run take the session id from onstat -u and run onstat -g ses <session-id> several times as see which query is taking the time. Are you transfering blobs across a slow network? If so the online poll threads could be spinning continuously trying to transfer the blobs across the network. This is a known Online problem - I'm not sure it is has been fixesd yet. If you still have problems, send us an onstat -a. -- David Williams