[Ocfs2-users] ocfs2 cluster becomes unresponsive

[Alexei_Roudnev]

It was THER SAME cluster. Just disk IO was unusial - it worked as
- delayed for 10 seconds (approx)
- then all IO comes thru
- then delayed again
- loop

As a result, cluster did not recognized IO problems and never saw a second
node (so conflicted for the heartbeat slot).

To make it work, we just fixed iSCSI config and recreated OCFS2 fs, but did
not touched cluster configuration.

----- Original Message ----- 
From: "Sunil Mushran" <sunil.mushran at oracle.com>
To: "Alexei_Roudnev" <Alexei_Roudnev at exigengroup.com>
Cc: "Andy Kipp" <AKIPP at velcro.com>; <ocfs2-users at oss.oracle.com>
Sent: Saturday, March 10, 2007 10:51 PM
Subject: Re: [Ocfs2-users] ocfs2 cluster becomes unresponsive


> The config error I would imagine would be that you defined two
> different clusters, each not having the other node, and that the two
> nodes have the same node number in both clusters. If so, the disk hb
> would have detected this error. It would have spewed error messages
> indicating that "some other nodes is heart beating in my slot". But yes,
> it would not have fenced.... well I'll need to read the code to confirm.
>
> Alexei_Roudnev wrote:
> > Just into your collection of _strange_ situations. I saw it few month
ago.
> >
> > We built 2 node cluster with iSCSI shared disks. Due to configuration
error,
> > servers got the same nodeID, and
> > it resulted in flip-flopping connection to the shared disk between
them -
> > fist server catched disk for 5 - 10 seconds, then
> > second catched disk, then first and so on.
> >
> > Result - OCFSv2 assigned the same node slot to both nodes, never
recognized
> > that other node was active, and
> > never fence or diagnose any problem (except uncyncronized IO, of course,
> > which broke file system in some time).
> > Looks as heartbeat alghoritm have some flow and don't detect some
failures.
> >
> >
> >
> > _______________________________________________
> > Ocfs2-users mailing list
> > Ocfs2-users at oss.oracle.com
> > http://oss.oracle.com/mailman/listinfo/ocfs2-users
> >
>