Many of our apps get “sick” before they die: they’ll miss 1-2 keepalives over a period of hours or days before they miss enough to be marked down. We’d like to know when they’re in that condition so we can address the problem before they croak.
Is there a way to see when things are “dying” but not yet dead? Either a counter on failed monitor checks or an info log message or something… I don’t see anything like that in the LOCAL-MIB.
Any suggestions?
Thanks - Al
Hi Al,
You could configure a script to run if a pool member is marked down and an alert is logged to /var/log/ltm. See this post for some additional info:
Generic command in user_alert.conf
DevCentral - An F5 Technical Community
And SOL9569:
SOL9469: A disabled interface is enabled after a reboot or an mcpd process restart
404 Page not found | BIG-IP Documentation
Aaron
Sorry… I just realised nothing is logged when a single monitor attempt fails--only when a monitor marks a pool member down. So the above suggestions won’t help much. You could add a second external monitor to the pool which runs a curl command and logs a message to /var/log/ltm on a single failure. You’d want to set the pool to mark the member up if at least one monitor succeeds so this second monitor won’t actually take the pool member out of the rotation. You could then trigger an SNMP trap, email or syslog message to a remote host for single failures. I’d be surprised though if you didn’t get a lot of false positives from temporarily busy pool members.
Aaron