[Ocfs2-users] Help tracing

[Angelo McComis]

Sunil/All

I recently came to understand through the help on this list how to  
stabilize my fencing problem. Read back through list archives this  
month for what I mean by that.

My concern now is that I have two possible fixes to my issue. (a)  
using datavolume,nointr,noatime for mounting my ocfs2 volumes and (b)  
raising the heartbeat threshold from 31 up to 61 or others I've seen  
have it as high as 76.

I have an IBM SVC storage environment and my storage folks show me  
graphs from their side showing no operation round trip from my side  
taking longer than 20msec on a bad day.  I contrast this against  
occasional scsi errors in netconsole logs and evidence of heartbeat  
timeout fencing.

I've looked at trace output from debugfs.ocfs2 -l HEARTBEAT allow, and  
all that seems to show is the 2sec hb flowing in the logs.

What should I be looking at to pinpoint and find the definitive set of  
tweaks that I need to push out?

Thanks. And thanks to everyone on the list who makes this the valuable  
resource that it is.



--
- Angelo
>