onstat -p all zero

Thanks for your reply Jonathan Leffler, I do have a performance problem, some users have complained of 'slows' with the OLTP part of the application. But this hasn't been quantified yet. Without an accurate definition of 'slows' I decided to investigate the performance anyway. The checkpoint duration is getting longer on average assessed over a period of 12 months with some infrequent but large values. Thats when I discovered that onstat -p behaves in the following way: After the database is taken offline and back up again, (done weekly) onstat -p shows some small static values that do not change regardless ofthe database use. I tested this by doing a load/query and the IO was reflected in the onstat -g iof stat's but onstat -p values were unchanged (accept for usercpy, syscpu and lchtwaits). Applying onstat -z sets the onstat -p values to zero, and they stay that way till the weekly recycle. The same thing happens on the development database. IBM Informix support suggested that I try re-installing Informix. There is an upgrade to IDS 7.31 on the way so I'll wait till then. Although this may not help. A comp.databases.informix search turned up another instance of this problem and surprisingly had similar platform : Dec 4100 alpha Tru64 Digital Unix v 4.0F running ids 7.31 ( same in this case accept we're running 7.24.FC6) Back to the performance issue: onstat -F has zero for LRU writes and Chunk writes and foreground writesalthough I observed the cleaners at work for the duration of the checkpoint and during this time vmstat reports 0 idle CPU resources available. There appears to be intense CPU usage during the checkpoint inteval (checkpoint inteval - 3 minutes). I was endeavouring to distrubute some of the IO between checkpoints using LRU writes to see if I could reduce the checkpoint duration and perhaps the CPU utilisation, but of course I need onstat -p. I found your reply in a comp.databases.informix search. Cheers Zanne Fagerstrom Zanne Fagerstrom wrote: > my problem is that onstat -p shows all zero's accept for usercpy, > syscpu and lchtwaits. Table sysptprof show's some small figures > which are too small to be correct. onstat -g iof look's correct. > > I thank you for any assstance on this matter. Jonathan Leffler replied It sounds like there's a problem, but we've far too little information to be able to come up with much in the way of a solution. Things like platform and DBMS version are very important. Is the database running OK or is it running too slowly, or is it down completely? Is there a performance problem? When was 'onstat -z' last run? Etc... I'd be inclined to try either restarting the database engine or running 'onstat -z' and see whether things alter. You can then run some simple SQL and see how the page reads and writes information changes. If it doesn't change, then you'd have to be suspicious that you're not monitoring the system you thought you were using. Zanne Fagerstrom