Forum Discussion

ijdod's avatar
ijdod
Icon for Nimbostratus rankNimbostratus
Jun 04, 2014

Node down after sync

Something I ran into after our (otherwise successful) upgrade to 11.4.1HF3. I was wondering someone else also ran into this issue.

 

Two LTM boxes, in sync, all pools and members are up(green). To test our NMS, I forced a (pool)member down (black) on the inactive LTM (LTM-B). The LTM reports out-of-sync, but I did not sync the configs yet at this point, as this is expected. I re-enabled the member. All members are now up/green again on both nodes, but the LTMs are out of sync. When I sync LTM-B to LTM-A, the member goes down/red on LTM-A, while remaining up/green on LTM-B. To fix this I have to sync LTM-B to LTM-A again with the 'overwrite configuration' flag enabled. This is reproducible on my system.

 

Disabling/enabling the member on LTM-A works (member comes up/green), but as this is a change that requires a sync, the above will happen again.

 

Update: this also happens if I sync the forced member down immediately afterwards, and again when I bring it back up. Also, disabling on node level works identical. Also, no log entries pointing at something suspicious, other than the node going down due to the monitor failing.

 

Further testing reveals this is tied to the 'incremental sync' feature. Turning that off makes the whole shebang work as expected.

 

3 Replies

  • Can you do this for me - On the ltm A when the pool member goes down can you run show /sys mcp-state and post the result. I am not sure at this point but I had a similar issue and refered this artical - http://support.f5.com/kb/en-us/solutions/public/13000/000/sol13030.html

     

  • When I sync LTM-B to LTM-A, the member goes down/red on LTM-A, while remaining up/green on LTM-B.

     

    does it go up/green if you leave it for a while? or does it stay down/red forever?

     

    have you tried tcpdump health monitor traffic on ltm-a when doing configsync?