Long checkpoint duration
Hi folks,
I do have a severe problem with checkpoint durations on
a Windows 2003 Server with Informix 9.40.TC6.
Well, it does not look like a performance problem, there
are only 5 to 8 users on that database and a script that
checks regularly Informix for current performance
information.
The system runs fine for days, checkpoint time is 0 secs.
But sometimes (I cannot define that more precisely) the
script hangs when requesting the current DBSpace usage
with this statement (which looks ordinary to me and runs
fine against many other Informix installations on Unix):
select name, sum(chksize), sum(nfree)
from sysdbspaces, syschunks
where sysdbspaces.dbsnum=syschunks.dbsnum
and is_sbspace = 0
and is_blobspace = 0
group by name
order by name;
"onstat -" then shows a CKPT REQ.
"onstat -u" then shows a 'C' at flag 1 for the corresponding
user thread.
The checkpoint may take several hundred seconds, even more than 30000.
Any ideas what is happening? What the problem can be?
Any, really any hint is highly appreciated!
R'diger
During the checkpoint, do you see activity from the flushers? Does
onstat -F show them active. Do you see write activity for those guyswith onstat -u? Do you see chunk writes with onstat -D?
What's your checkpoint interval? lru max dirty? lru min dirty? How
big is your phsical log?
Up to 8 hours for a checkpoint sounds like a performance problem to me.
select statement hangs too?
select is unlikely to be affected byt he checkpoint unlesthere is a
lock contention issue
sounds like you may have a hardware problem on one of your disks
Keith wrote:
> During the checkpoint, do you see activity from the flushers? Does
> onstat -F show them active. Do you see write activity for those guys> with onstat -u? Do you see chunk writes with onstat -D?
>
> What's your checkpoint interval? lru max dirty? lru min dirty? How
> big is your phsical log?
>
> Up to 8 hours for a checkpoint sounds like a performance problem to me.