Pre 1.0.12, the heartbeat thread did heartbeating and handled requests
for disk voting. On rare occasions, the disk voting would delay
heartbeating enough for other nodes to drop that node from the cluster.
Not a good thing.
In 1.0.12, the heartbeat thread does heartbeat and submits a task for
the disk vote requests so as not to delay the hb.
So, if you are on 1.0.11 or earlier, then upgrade to at least 1.0.12, if
not 1.0.13.
However, if you are on 1.0.12+, then look for any errors on your system
which is preventing the hb thread to write to the disk. Like SCSI
errors. See if there is any relationship to this and the io load on the
box at that time.
On Thu, 2004-10-07 at 15:37, David McWhinnie wrote:> Can anyone give me some information about what this
> error really means.
>
> Oct 7 15:49:23 houided001 kernel: ocfs: Removing
> houided002 (node 1) from clustered device (8,192)
> Oct 7 15:49:24 houided001 kernel: ocfs: Adding
> houided002 (node 1) to clustered device (8,192)
>
> David.
>
>
>
>
>
>
> __________________________________
> Do you Yahoo!?
> Yahoo! Mail - You care about security. So do we.
> http://promotions.yahoo.com/new_mail
> _______________________________________________
> Ocfs-users mailing list
> Ocfs-users@oss.oracle.com
> http://oss.oracle.com/mailman/listinfo/ocfs-users