I’m trying to write an iRule that will down a member in LB_FAILED. I figured this would be a good spot to do it, since LB_FAILED is reached if a selected member is unreachable or non-responsive.
However, it seems that I can’t get information for the selected member. [LB::server pool] returns properly, but [LB::server addr] and [LB::server port] don’t give me anything, and I therefore can’t call LB::down with the appropriate arguments. Calling LB::down alone doesn’t seem to do anything in this context either.
I think this is because of the way LB_FAILED works (someone please jump in if I am wrong): it’ll fire if the LTM can’t select a pool member, or that member is down. So to me it’s not a surprise that you can’t down a member from this event.
What are you trying to accomplish, specifically? That may help home in on some options.
I would have expected that LB::server commands would return info for the “currently selected” pool member even if the connection failed. If a member couldn’t be selected because no pool member was up, then it would seem logical that LB::server * could only return the pool name. I could see why the IP::server_addr/TCP::server_port commands wouldn’t return anything in LB_FAILED because no connection was established to the pool member.
Maybe this is a bug? There have been quite a few issues with using commands in LB_FAILED over the past versions. However, from a post post, someone was able to at least get the LB::server addr and LB::server port values in LB_FAILED (in 9.3.x):
Ahhh yes very good points Aaron, thanks. Now I’m going to have to go off and research this one…so the last question is key: are there any members available? If so, we’ll need to dig into this one a bit!
Ok my apologies everyone. The problem is that I was doing the LB::detach before the LB::down. The detach will “unselect” any currently selected member. Makes sense.
It’s just a basic failover iRule. The idea is to use a designated “failover pool” for a given pool. I know I could do this with priorities but I need to snat as the “failover pool” will be a vip on an LB at an alternate datacenter. In addition to adding the snat, I inject a header to try and prevent any crazy loops between LB’s. If you have any suggestions or improvements, especially regarding efficiency, do share.
when RULE_INIT {
set ::remoteDCSnatPool "remote-dc-snat"
}
when CLIENT_ACCEPTED {
set defaultPool [LB::server pool]
set failoverPool "[LB::server pool]-failover"
set failover 0
}
when HTTP_REQUEST {
if { [active_members $defaultPool] < 1 || $failover == 1 } {
if { [HTTP::header exists lb_origin] } {
log local0. "lb_origin detected, rejecting"
reject
} else {
log local0. "Failing over to $failoverPool"
set failover 1
HTTP::header insert "lb_origin" 1
snatpool $::remoteDCSnatPool
pool $failoverPool
}
}
}
when LB_FAILED {
if { $failover == 1 } { reject }
LB::down
LB::detach
log local0. "reached LB_FAILED, [active_members $defaultPool] active members"
if { [active_members $defaultPool] < 1 } {
if { [HTTP::header exists lb_origin] } {
log local0. "lb_origin detected, rejecting"
reject
} else {
log local0. "Failing over to $failoverPool"
set failover 1
HTTP::header insert "lb_origin" 1
snatpool $::remoteDCSnatPool
LB::reselect pool $failoverPool
}
} else {
LB::mode rr
LB::reselect
}
}