[Ocfs2-users] fencing timeout- increasing min time
Christopher Duncan
duncancj at comcast.net
Thu Sep 29 01:43:30 CDT 2005
What would the dangers/issues be within increasing the minimum time
before the fencing enforced panic of a node? We're hitting problems where
HBA multipath failover does not happen fast enough and we may need to up the
fencing timeout to something closer to 30-60 secs or more. Is this mostly a
performance impact or does it increase the window for corruption issues?
Looks like from the current code the min timeout is 4 secs.
For hardware/software details we're using CentOS4.1 x86_64 (clone of
RedHat 4.1),
on a cluster of Sun v20z's, Opteron 250s. The cards involved are Emulex
and multipath
is mdadm based.
More information about the Ocfs2-users
mailing list