Ocfs2 users - Feb 2009 - No space left in device

Hello,

We had an issue today in one of our servers. We were unable to create more 
files in some of the nodes. On the others it was possible.

We were using:

Feb 11 12:46:51 t1 kernel: OCFS2 Node Manager 1.2.6 Fri Sep 28 10:10:03 PDT 
2007 (build 1c4508be0c999c7bec8d997b52240db8)
Feb 11 12:46:51 t1 kernel: OCFS2 DLM 1.2.6 Fri Sep 28 10:10:03 PDT 2007 (build 
1c210e3c0b5b465c1ea15c7d01819c19)

Aparently the problem we had is the one i already reported a few months ago:

http://oss.oracle.com/bugzilla/show_bug.cgi?id=1023

We've upgraded all the cluster to 

Feb 11 12:50:12 t1 kernel: OCFS2 DLM 1.4.1 Tue Dec 16 19:18:05 PST 2008 (build 
4ce8fae327880c466761f40fb7619490)
Feb 11 12:50:12 t1 kernel: OCFS2 DLMFS 1.4.1 Tue Dec 16 19:18:05 PST 2008 
(build 4ce8fae327880c466761f40fb7619490)

And the problem disapeared. I would like to ask if that problem was solved in 
the 1.2->1.4 code, or will we have this problem later? If so, what should we 
do to prevent it?

Thanks for any info,
Best regards,
Nuno Fernandes
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://oss.oracle.com/pipermail/ocfs2-users/attachments/20090211/86c79a1d/attachment.html

bug#6802646
The same fix is available in 1.2.9-1.
http://oss.oracle.com/projects/ocfs2/news/article_18.html

The bugzilla#1023 describes the problem. The above fix helps
but does not fix the core issue entirely. We are working on
further improving inode allocations. But the reality is, as
long as we allocate inodes dynamically, we will have to live
with the fact that there will always be a chance that the disk
is not 100% used but there is not enough contiguous space to
grow the inode allocator. One easy solution is to pre-allocate
inodes during mkfs, which is typical in most filesystems. There
it leads to space wastage because not all inodes are used.

Sunil


Nuno Fernandes wrote:
>
> Hello,
>
> We had an issue today in one of our servers. We were unable to create 
> more files in some of the nodes. On the others it was possible.
>
> We were using:
>
> Feb 11 12:46:51 t1 kernel: OCFS2 Node Manager 1.2.6 Fri Sep 28 
> 10:10:03 PDT 2007 (build 1c4508be0c999c7bec8d997b52240db8)
>
> Feb 11 12:46:51 t1 kernel: OCFS2 DLM 1.2.6 Fri Sep 28 10:10:03 PDT 
> 2007 (build 1c210e3c0b5b465c1ea15c7d01819c19)
>
> Aparently the problem we had is the one i already reported a few 
> months ago:
>
> http://oss.oracle.com/bugzilla/show_bug.cgi?id=1023
>
> We've upgraded all the cluster to
>
> Feb 11 12:50:12 t1 kernel: OCFS2 DLM 1.4.1 Tue Dec 16 19:18:05 PST 
> 2008 (build 4ce8fae327880c466761f40fb7619490)
>
> Feb 11 12:50:12 t1 kernel: OCFS2 DLMFS 1.4.1 Tue Dec 16 19:18:05 PST 
> 2008 (build 4ce8fae327880c466761f40fb7619490)
>
> And the problem disapeared. I would like to ask if that problem was 
> solved in the 1.2->1.4 code, or will we have this problem later? If 
> so, what should we do to prevent it?
>
> Thanks for any info,
>
> Best regards,
>
> Nuno Fernandes
>