mkyrc
Sep 17, 2024Cirrus
GTM pool is OFFLINE even if pool members are UNKNOWN
Hi,
Maybe someone can clarify me this situation. I didn't found it in documentation.
- generic host (with no monitors) has two virtual servers (also no monitors here). the state of server and virtual servers is unknown (that expected).
- state of the pool is OFFLINE (why?? this is not clear for me), but (all two) members are UNKNOWN
- wide IP is OFFLINE because poll has no available members (members are unknown, not unavailable)
- dns response to wide ip returns two IPs (IP adresses of both members).
- it's ok in this case, because return code on failure is (by default) disabled
- when I enable 'return code on failure', response is empty
- Note: when one member is disabled (or down based on temporary monitor), dns response return only one IP - IP of the unknown member. That's correct, but pool state and wide ip state are offline.
My question is: Why is pool state OFFLINE when pool members states are UNKNOWN?
I think he should be unknown. When the same situation occurs on LTM, state of pool is unknown, not offline. Does GTM behave differently??
TMOS version: 17.1.1.3
Here is simple test configuration:
# gslb domain (wide ip)
gtm wideip a /testTenant/testApp/test.my.local {
pools {
/testTenant/testApp/testPool { order 0 }
}
}
# gslb pool
gtm pool a /testTenant/testApp/testPool {
alternate-mode global-availability
fallback-mode none
load-balancing-mode global-availability
members {
/Common/server1:vs1 { member-order 0 }
/Common/server1:vs2 { member-order 1 }
}
}
# gslb servers
gtm server /Common/server1 {
datacenter /Common/testDc devices {
0 {
addresses { 10.1.1.1 { } }
}
}
prober-fallback none
product generic-host
virtual-servers {
vs1 { destination 10.1.1.11:0 }
vs2 { destination 10.1.1.12:0 }
}
}
I found article describing this behavior (https://my.f5.com/manage/s/article/K10302440). It answers my question, but it's still confusing to me.