Forum Discussion
Frank_Mancini_3
Nimbostratus
Nimbostratususing active_members <poolname> takes a long time to failover
Hello everyone. I have a quick question
I'm using
when CLIENT_ACCEPTED {
if {[active_members my_pool] < 1 } {
pool new_pool
persist tcp
}
Now, when i shutoff my 3 members I have in the "my_pool" (all of them), it seems I get a "Server Not Found" error for about 15 seconds as I click refresh. After the 15 seconds "new_pool" seems to kick in and I get that pool.
Can someone explain this? Is this a function of my monitor that I have applied to this pool? What are the affects if i set the threshold lower?
Thank you so much for any replies!
Franko
- Hi Franko,
I think you made a good call there: If by shutting off the members you mean that you disable the service on the servers themselves, it is most likely a function of the monitor timeout, which by default is 16 seconds. (You can verify that's what you're seeing by disabling the pool members in the LTM config, then the rule action should immediately reflect 0 pool members, without the 15 second delay.)
Setting the monitor timeout lower is possible, but not necessarily recommended, as you'd also have to set the interval for the checks correspondingly lower, at which point you may be unintentionally DoS'ing your servers with health checks.
/deb 
Frank_Mancini_3Thank you.
Are you saying that when you shut down one of the members in the pool, you get the same error on your BigIP?
I get
"Server Not Found"
for those 15 seconds.
-franko- Well, if you happened to shut off the member exactly 1 second after the monitor last polled, then you could possibly get a "server not found" response for the remaining time until the monitor fires again, yes. This would, of course, require that you're persisting to the same member, but it is theoretically possible.
Colin
