Ryan Kish wrote:
I am under the impression that using TCP would cause a TIME_WAIT issue
during node failover. Its actually stated at hint #2 on the Linux-HA
website's NFS page: http://www.linux-ha.org/HaNFS
If this is incorrect, I am more then happy to switch over to TCP!
You can most certainly use TCP, I've done it and it works... That said, I still
have yet to thoroughly test various conditions of failover and you should take
this advice with a grain of salt.
If I'm reading it correctly, hint #2 is describing a particular situation where
failover happens from node A to node B then back to node A right away. In this
case, the failover will by sloooow with TCP because of the TIME_WAIT issue.
However, if you just want failover from node A to node B then it works fine,
albeit not instant failover. The whole process from incident on the cluster,
through failover and finally to client recovery takes about 2 minutes in my
setup. This of course is related to deadtime and also how many resources need
to failover. Since I have a few dependencies during failover, I'd say half of
that 2 minute time is starting/stopping resources.
--Ryan Thomson
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
