Re: Spurious failovers

On Thu, 2014-11-20 at 18:03 +0000, John Horne wrote:
> Well in the end we decided to run an strace on the keepalive processes
> overnight. The problem occurred again last night, and the culprit seems
> to have been within a 'select' statement:
<snip>

Are the affected machines VMs? (goes to read thread on CentOS list...)
Ah, yes they are.

Check your virtualisation platform. This smacks of a VM snapshot being
taken by a backup system, or the VMs being moved from host to host for
maintenance or system load reasons.

We have a big VMware platform here and I've had to turn down our HA
sensitivity (both in keepalived and heartbeat, where used) because
snapshots can cause the VMs to "pause" for a few seconds when they are
very near completion; the VM state is quiesced before the task
completes. On most quiet systems this is not noticeable but on systems
with a high rate of data change or time-sensitive processes it can be a
problem, causing as you've seen spurious failovers.

Graeme


------------------------------------------------------------------------------
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server
from Actuate! Instantly Supercharge Your Business Reports and Dashboards
with Interactivity, Sharing, Native Excel Exports, App Integration & more
Get technology previously reserved for billion-dollar corporations, FREE
http://pubads.g.doubleclick.net/gampad/clk?id=157005751&iu=/4140/ostg.clktrk