Forum Discussion
Guillaume_Rouss
Nimbostratus
Nimbostratusdebugging LB_FAILED event
Hello list.
I'm using this irule in order to deliver a user-friendly error page to users when a service is down:
when LB_FAILED {
log local0.debug "LB_FAILED for [virtual name]"
snatpool server_snat_pool
LB::reselect pool fallback_https_pool
persist none
}
Following suggestion from https://devcentral.f5.com/s/feed/0D51T00006i7Q1cSAE, I added a call to [event info] in the log message, in order to get a better idea of unexplained such events. However, this isn't really useful, as:
- 50% of log messages have no reason
- 50% of log messages have "connection limit" as reason, whereas I don't have any explicit limit defined at virtual server level, pool level or node level
So, is there any other place I should check for connection limits, and is there any better way to identify LB_FAILED event reasons ?
- Lucas_Thompson
Employee
The connection limit sounds like a good hint. You might want to keep an eye on the number of flows to that vip using FLOWTABLE::count to log that also.
Maybe you could count the connections on that VS and in whatever TMM is used for that flow:
log local0. "LB_FAILED! info: [event info] , current conn in this TMM: [FLOWTABLE::count], current conn in vs [virtual name]: [FLOWTABLE::count virtual [virtual name]]"
Do you see any other /var/log/LTM error logs like SNAT pool exhaustion or the like?
