Symptoms

TMM may core if an APM virtual server detaches from its current resource and attaches to a different resource while handling requests.

Impact

Traffic disrupted while tmm restarts.

Conditions

This crash is most likely to occur when an APM virtual server is configured with a OneConnect profile. However, as long as there is the possibility that the resource APM should connect to is decided after the client connection is established (e.g., based on the HTTP Host header), this issue is possible. This would include iRules that change the backend resource, the load balancer makes a decision to switch the resource, or APM is configured to possibly interface with a number of different resources.

Workaround

If OneConnect is present, removing its profiles from APM virtual servers greatly reduces the likelihood of this issue occurring. If the issue occurs without OneConnect, the other mitigation is to place APM in its own virtual, configured to interface with a layered virtual which remains constant. The layered virtual may then use iRules or load balance as normal. The APM virtual server cannot detach from one resource and re-attach to a different resource.

Fix Information

APM virtual servers that can cause the resource to switch during request handling (as is most noticed with OneConnect profiles attached to them) will no longer cause TMM to crash and restart.

Behavior Change