ACE Rserver "inservice" - probing?

[graemerobertson-dc]

Hello all

We recently upgraded an ACE module from A1 code straight through to A2(3.4)

Theres a behaviour change we werent expecting -

probe tcp t2-probe-3133

  port 3133

  interval 4

  faildetect 4

  passdetect interval 4

  passdetect count 4

  receive 1

serverfarm host ext-gxr-3133

  probe t2-probe-3133

  rserver server-testing 3133

  inservice

If I take the "rserver server-testing" out of service and then bring it back in, it goes straight to OPERATONAL even if the service listening on port 3133 is not there - the probe eventually fails, the server drops to OUTOFSERVICE.

During this time we drop transactions on the serverfarm.

This is different from the A1 behaviour - it used to wait till the probe finished (We're pretty sure anyway :-) )

Am I dreaming - this hasnt changed ? - And regardless, is there a way to make the behaviour "Wait until the probes work before bringing the server in" ?

Cheers

Graeme.

[chrhiggi]

Graeme-

  I just tested this on A1(6.3) and A2(3.3) - both do exactly the same in terms of thier default action.  When the rserver is operational with no probe configured, and you add a probe, the rserver stays operational until the probe fails.  If the rserver was in probe failed state to begin with and you add a 2nd probe to the serverfarm, the rserver stays in probe-failed while testing the the new probe. 

There was a difference in A1(6.X) vs A2(3.X) - the addition of "fail-on-all" under the serverfarm which makes all probes have to fail prior to removing it from loadbalancing rotation.  I did test with that feature on, and it still has the same result.

If you are getting something different from this, go ahead and get a TAC case open to have a bug investigated/filed.

Regards,

Chris Higgins