Inband monitor not working correctly
I am testing applicability of Inband monitor in my environment.
I was testing how it is handling members (servers) that are slow - are replaying after configurable timeout.
I noticed that Inband monitor is marking member down, once it discover 3 responses sent after about 7-10 millisec.
According to default configuration it should mark member down if it sent response after more than 10 sec.
Question: What may be causing this incorrect monitor behavior? How I should configure it so it will work as expected?
I am using Virtual Server Type: Performance Layer 4. I have not changed other settings.
I have pool with default inband monitor assigned.
I noticed this on virtual BIGIP-10.1.0.3341.1084 (VE),
and hardware device running BIG-IP 10.0.1 Build 402.7 Hotfix HF4.
I am attaching sample of successful exchange of packets between BigIP and Server (member). With this exchange F5 was keeping member marked as up.
I am also attaching sample of unsuccessful exchange of packets, that leads to marking member as down.
The only difference is, that on successful exchange sample, time difference between data packets [PSH/ACK] from client (bigip) and server (member) is ~1msec.
In unsuccessful sample this difference is ~101msec. (as I stated ablow I noticed this wrong behavior when this difference was larger than ~10msec)
BTW, when I changed Virtual Server Type to Standard, Inband monitor has never marked member as down, even when it was responding after more then 10 sec. Which again is not a correct behavior.