geom_raid5 livelock?
R. B. Riddick
arne_woerner at yahoo.com
Fri Jan 12 20:36:59 UTC 2007
--- CyberLeo Kitsana <cyberleo at cyberleo.net> wrote:
> Hence the testing. I wish to ensure there are as few problems as
> possible when this is put into production.
>
OK... I understand that... Testing is important, but the programmer tends to
test not so good... ;-)
> The kernel does not panic, and everything resumes just fine as soon as
> safeop is disabled. There are no new messages in the kernel log, nor in
> syslog, and all disks are operating properly, if slowly (around 40
> kilobytes per second each, with dd bs=4096).
>
Hmm - and at the same time u feel that the processes have to wait much longer
than necessary? Or is it possible, that the processes r just idle and read not
so often? What does "gstat" say (a screen shot during a suspected live lock
would be fine)? CPU usage according to "top"?
> Also, gcache doesn't seem to exist in the base system. Is this easier to
> build than gjournal? (patching for that caused a whole ton of kernel and
> world problems I'd rather not revisit)
>
Hmm - gcache is possibly only in 7-CURRENT...
> The attached list was taken during the most recent lockup.
>
Looks perfect to me... :-)
You could try to activate graid5's debugging by setting the debug level to 2:
sysctl kern.geom.raid5.debug=2
Then u should see the requests in /var/log/messages...
I would be interested then in those messages during a suspected live lock (just
wait for that strange condition and then set the debug level to 2 for some
seconds until u feel u have enough data...)...
-Arne
____________________________________________________________________________________
Do you Yahoo!?
Everyone is raving about the all-new Yahoo! Mail beta.
http://new.mail.yahoo.com
More information about the freebsd-geom
mailing list