GJournal too Small?

[Ivan Voras]

Tony Holmes wrote:
> I have a 894GB gstripe that I've put gjournal on top of. Since it was
> an unused stripe I placed the data+journal into the same partition.
> The server is fairly heavily used and hung suddenly. On reboot I got
> this in the dmesg:
> 
> GEOM_JOURNAL: Journal 3472355975: mirror/gm0s1e contains journal.
> GEOM_STRIPE: Device st0 created (id=2649322337).
> GEOM_STRIPE: Disk mirror/gm0s1f attached to st0.
> GEOM_MIRROR: Device mirror/gm1 launched (1/1).
> GEOM_STRIPE: Disk mirror/gm1s1f attached to st0.
> GEOM_STRIPE: Device st0 activated.
> GEOM_JOURNAL: Journal 2630378703: stripe/st0 contains data.
> GEOM_JOURNAL: Journal 2630378703: stripe/st0 contains journal.
> GEOM_JOURNAL: Journal stripe/st0 clean.
> GEOM_JOURNAL: Timeout. Journal gjournal 3472355975 cannot be completed.
> 
> That last line worries me.
> 
>>From a quick google, it appears that the gjournal is too small. Since 
> I created it with the single partition, I would have expected the journal
> to be autosized correctly. 

The message "Timeout. Journal %s cannot be completed." is printed when 
the gjournal composite device is created with data and journal on 
separate providers. It means that a timeout occurred while gjournal 
waited for both providers to come online. Your message contains 
something that looks like a journal ID (3472355975) which isn't in the 
code in 7-stable and 8-stable/head. How did you get that line?

Assuming magic has happened and the journal ID (3472355975) is correct 
then it means you have two gjournal devices, one of those created on the 
stripe st0 (2630378703). It could mean that somehow, there is still 
recognizable metadata on your drives and/or partitions which confuses 
gjournal.


> I know the OS is a little out of date but has been working very well until
> the past couple months. I have 1 hang approximately every month.
> 
> Information about system:
> 
>   FreeBSD fs.cwahi.net 7.1-PRERELEASE FreeBSD 7.1-PRERELEASE #0: Mon Dec  1 09:12:42 EST 2008 root at app.cwahi.com:/usr/obj/usr/src/sys/CWahi  amd64
> 
>   fs# gjournal list
>   Geom name: gjournal 1493846988
>   ID: 1493846988

This, on the other hand, is a third gjournal ID.

Assuming that somehow all this information is correct, you should 
probably send the output of sysctl -b kern.geom.confxml before anyone 
can unravel what has happened :)