Forum Discussion
NimbostratusCan we log why a client gets bounced to different backend servers?
Developers noticed their instance got bounced as they were working, but should be not as long as the cookie is set.
We have F5 default cookie enabled with longer timeout 3600.
We have also IP Tracking enabled for uploads and WebDav connections.
Developer was just interacting with the site clicking pages when suddenly prompted to log in again. But noticed clearly it was due to backend server change, as timeout was too short for cookie expiration. Just a minute or two between clicks.
I could not find a way to view reasons for expiration of cookies or other reasons to switch client to a different host.
I could find the persistence table, but it seems like the active connections, without a reason.
show /ltm persistence persist-records
Maybe I hit it wrong from the ltm though.
Any help to log the reason for re-connecting to different backend would be great.
Leonardo_SouzaCirrocumulus
There is no way to log why there was new load balancing decision. You can check the logs around that time, and if the pool member was down, you can easily conclude the new load balancing decision was because of that.
You could have an iRule logging a lot of information for every request, but that is not the correct way.
The correct way is to take a tcpdump while you replicate the problem, and analyze that after.
Also, important to mention that a load balancing decision occurs per connection by default, and not per HTTP request.
This solution explains the difference between load balancing per connection or per request.
https://support.f5.com/csp/article/K7964
If you want to play with tcpdump, here is the solution:
https://support.f5.com/csp/article/K411
- P_K
Altostratus
Applying oneConnect profile to your VIP should solve the problem.
