Re: Reducing checkpoints

On 19 Jun 2006 13:13:25 -0700, "PapaKiKi" <agwanzura@okzim.co.zw> wrote: >Hi,i have a database with 80gigabytes of data with 360users and up to >400 when at peak.I use IDS 7.31 running on HP-UX 11.0 and 8 Gig of >memory.The checkpoints get to as high as 58sec at peak periods .Users >often run out of connections into informix.How can i reduce these >checkpoints to acceptable levels like 3seconds.What others measures can >help me.Below is the configuration from the onconfig.live in informix 360-400 users . . . running out fo connections. 58sec checkpoints at peak periods .. I'm presuming that these are pretty consistent . .same time of day? BTW, the onstats output that others have recommended would be a good idea. Good to have some hard data to see what's going on. >ROOTSIZE 1000000 # Size of root dbspace (Kbytes)Seems a bit big . . . guess it doesn't really matter anymore, especially if the physical log is in the rootdbs. >LOGFILES 30 # Number of logical log files >LOGSIZE 5000 # Logical log size (Kbytes) Any long transactions? How often do your logical logs fill? >NETTYPE ipcshm,1,800,CPU # Configure poll thread(s) for nettypeAre your users connecting through shared memory or TCP? My guess is shared memory. You're configured for 800 shared memory connections, which are locked in . . the 801st connection will be refused. I'd recommend something along the lines of . . NETTYPE ipcshm,4,250,CPUI'm not sure the maximum number of connections a poll thread should handle . .. 250 to 300 . . >BUFFERS 200000 # Maximum number of shared buffersAnything else running on this box? 400M of buffers seems low. I'm curious how often the buffers are turning over. >NUMAIOVPS 8 # Number of IO vpsAre you using KAIO?? How many chunks do you have? >SHMVIRTSIZE 300000 # initial virtual shared memory segmentAny additional shared memory segments? What does onstat -g seg say? >CKPTINTVL 360 # Check point interval (in sec)Not too bad, I'd guess . . . how far apart do these 58sec checkpoints occur? Every 6 minutes? >LBU_PRESERVE 0 # Preserve last log for log backup Might want to change this to 1. Informix may get into a situation which it can't get out of . .. . refer to the manual for further explanation. >RA_PAGES # Number of pages to attempt to read>ahead >RA_THRESHOLD # Number of pages left before next>group > Anyone know what the defaults are for this version? >DBSPACETEMP tempdbs # Default temp dbspaces Only 1 temp dbspace . . . how big? Just a few ideas . . . JWC