Virtual server interution when all nodes thrashing

We had an issue where all the nodes in a virtual pool were going up and down like yo-yos, and at one point they all went down, and then they all came back up a moment later... and then... NOTHING. It appeared that no traffic got through the virtual server to any of the nodes for over two hours. Then we disabled the virtual server, re-enabled it and POW! Traffic was passing.

 

 

Has anyone seen this before... Is there some kind of bug or a limitation on the number of times nodes can cycle before the bigip decides the virtual server is toast?

 

 

Is there anything we can do do smooth out the frequency that nodes are dropped, say a less frequent and more complex availability check instead of a simple port tcp monitor.

 

 

Thanks,

 

 

-Taylor