Re: checkpoint problem

scottishpoet wrote: > is this a new problem or has it always occurred? > > have you changed anything in terms of your workload that has caused > this problem to occur? > > the checkpoint is the time taken to flush data from memory to disk, is > there some ocntention on some devices that is causing this to take a > long time, too many dbspaces table son one device, what does sar report > as the disk activity across your devices during the checkpoint? > > do you possibly have a problem with one of your disks running slow? try > if possible dding to each of the devices and see what the performance > is like We have a similar problem with an AIX box after a hardware/OS/sortware upgrade (4-way p630/AIX 5.3/Informix 7.31UD8). When we changed bact to cooked FS we GAINED performance. A normal ontape -r of a 70G image backup runs about 30 minutes. With the raw FS it takes up to 6 hours. When creating the files on the cooked chunks with the dd command, the system pushed about 48M per second to disk. While the ontape ran it was about 40M/sec (still very good). During normal processing with only one user active it is1M/sec or less. Something is not right with this version of the Informix engine. I see no problems of this type in the UD9 documentation but am willing to try it.

> We have a similar problem with an AIX box after a hardware/OS/sortware > upgrade (4-way p630/AIX 5.3/Informix 7.31UD8). When we changed bact to > cooked FS we GAINED performance. > > A normal ontape -r of a 70G image backup runs about 30 minutes. With > the raw FS it takes up to 6 hours. this sounds typical of the sorts of problems if you confuse block and character special devices when you use raw devices