Re: [SSI-devel] ha-lvs hanging after ~24 hours ?
Brought to you by:
brucewalker,
rogertsang
From: <ia...@g7...> - 2005-10-11 21:16:14
|
On Mon, Oct 10, 2005 at 06:11:23PM -0400, Roger Tsang wrote: > Okay sounds like LVM2 lockup then. It can sometimes lock up all of a sudden > without warning. You can determine whether it is LVM2 or not. Try running > LVM2 on only one node and reboot the LVM2 node (that owns the original LVM2 > device) when this lockup happens. If the remaining nodes in the cluster > recover after rebooting the LVM2 node, then we know what the problem is. > > Obviously you would want to test this by running LVM2 on a non-initnode - > rebooting the initnode would kill the cluster since 1.9.1 doesn't failover > properly. Okay, I've setup one of my nodes with a VG, and created an lvol on it. I'm currently running a bunch of dd's (on the node in question) in parallel, and serially, in a loop, in an attempt to help speed up the lockup getting triggered. Oddly enough, the cluster has now been up for almost 4 days, and hasnt locked up (typical!). I'm just hoping the non initnode dies first... Iain |