gjournal questions
Kevin Kramer
kramer at centtech.com
Fri Sep 1 12:23:03 PDT 2006
i was seeing up to 14s suspends on /scr10 prior to the reconfig causing
mount point to be unresponsive on the clients. i'm using async option
also. the rsync is still going so won't go live again 'til next week.
thanks for the help.
------------------------------
Kevin Kramer
Sr. Systems Administrator
512.418.5725
Centaur Technology, Inc.
www.centtech.com
Pawel Jakub Dawidek wrote the following on 09/01/06 13:56:
> On Fri, Sep 01, 2006 at 10:16:09AM -0500, Kevin Kramer wrote:
>
>> I've already redone the whole thing. here are the steps i took
>>
>> umount /scr09
>> umount /scr10
>> gjournal stop da2.journal
>> gjournal stop da4.journal
>>
>> ** had not done this on the first attempt
>> newfs /dev/da1
>> newfs /dev/da3
>>
>
> This is not needed.
>
>
>> gjournal label -v /dev/da2 /dev/da1
>> gjournal label -v /dev/da4 /dev/da3
>>
>> newfs -J -L scr09 /dev/da2.journal
>> newfs -J -L scr10 /dev/da4.journal
>>
>> mount /scr09
>> mount /scr10
>>
>
> Don't know how your /etc/fstab looks like, but you definiately should
> use 'async' mount option, which is safe to use with gjournaled file
> systems.
>
>
>> it is looking much better so far. before, i was getting this in the debug (only on /scr10) and my mountd process was always the top process
>>
>> Sep 1 00:00:11 donkey kernel: fsync: giving up on dirty
>> Sep 1 00:00:11 donkey kernel: 0xc9ee1bb0: tag devfs, type VCHR
>> Sep 1 00:00:11 donkey kernel: usecount 1, writecount 0, refcount 198 mountedher
>> e 0xc9ebfb00
>> Sep 1 00:00:11 donkey kernel: flags ()
>> Sep 1 00:00:11 donkey kernel: v_object 0xc9fbb528 ref 0 pages 5933
>> Sep 1 00:00:11 donkey kernel: lock type devfs: EXCL (count 1) by thread 0xc9bd4
>> 180 (pid 38)
>> Sep 1 00:00:11 donkey kernel: dev ufs/scr10
>> Sep 1 00:00:11 donkey kernel: GEOM_JOURNAL: Cannot suspend file system /scr10 (
>> error=35).
>>
>
> It happens sometimes under load, haven't investigated yet what exactly
> is happening, but you can ignore it for now, it's harmless, it just
> means journal switch will be done a bit later.
>
> BTW. 8GB for journals is much. You should not need more than 2GB
> probably. Of course it will work with 8GB just fine.
>
>
More information about the freebsd-stable
mailing list