<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
</head>
<body bgcolor="#ffffff" text="#000000">
<font size="-1"><font face="Helvetica, Arial, sans-serif"><br>
Hi,<br>
<br>
I have 3 nodes cluster using OCFS2 1.4 on CentOS5.5 (kernel
2.6.18-194). Two nodes (server1 and server2) are providing shared
storage using DRBD. Shared storage exported to nodes using iSCSI
(server1 is target and all other nodes are iSCSI initiators).<br>
<br>
Today cluster went down. Server 1 was not accessable, while server2
and server3 got rebooted with log showing that connected to server 1
was lost, and because of that, servers were rebooted. However, in
server1's logs doesn't have errors like that. There are only following
lines minute before other two servers rebooted:<br>
<br>
Mar 18 20:58:03 server1 kernel:
(dlm_thread,5154,3):dlm_drop_lockres_ref:2216 ERROR: while dropping ref
on BDB600C633D74D6B85C496D78F566879:O0000000000000002e81a8700000000
(master=1) got -22.<br>
Mar 18 20:58:03 server1 kernel: lockres:
O0000000000000002e81a8700000000, owner=1, state=64<br>
Mar 18 20:58:03 server1 kernel: last used: 4501088944, refcnt: 3, on
purge list: yes<br>
Mar 18 20:58:03 server1 kernel: on dirty list: no, on reco list: no,
migrating pending: no<br>
Mar 18 20:58:03 server1 kernel: inflight locks: 0, asts reserved: 0<br>
Mar 18 20:58:03 server1 kernel: refmap nodes: [ ], inflight=0<br>
Mar 18 20:58:03 server1 kernel: granted queue:<br>
Mar 18 20:58:03 server1 kernel: converting queue:<br>
Mar 18 20:58:03 server1 kernel: blocked queue:<br>
<br>
I think that I saw, from time to time, errors like this logged in
/var/log/messages, but servers continue to work without hanging. Is
this kind of error serious enough for server to go down? If it is, why
is it happening and how to prevent it?<br>
<br>
Thanks,<br>
Nikola<br>
<br>
</font></font>
</body>
</html>