We were "forced" to upgrade from stable R80.30 T215 based on the fact that log search was totally screwed and fix in R80.30 would take long time to develop (see SK168616)
Was a very long night somewhat based on events outside CP (had to clean up VM snapshots and that took only 4hrs..)
Some of my observations that's worth sharing:
- Upgrade documentation for MDS HA using CPUSE is inaccurate for HA part as it actually tells you to install, not upgrade using CPUSE in the procedure and some parts are missing. So if you are not familiar with MDS HA upgrades, probably worth asking CP support to clarify (I don't want to make this post too long with instructions)
- Both Primary and Secondary MDS failed upgrade with the first attempt! It would restart against R80.40 and then soon after roll back itself to R80.30. Without doing a single change, I re-run upgrade using CPUSE and both succeeded with the second attempt. Very strange as both MLMs worked flawlessly. I had no time to investigate as night was not long enough and we desperately needed log search working for operations
- Pls make sure that you read sk163814 as I did rookie mistake by installing T77 on primary MDS before I upgraded secondary and MLMs, so they all stopped half way through the upgrade, until I uninstalled Take77 on primary.. This has worked before, even with R80.30 but not anymore. SK is really good!
- I noticed that backup size has shrunk by 20%, that's good news as R80 and R80.10 were really bad, when our backup went up to 25GB and now it's 7GB!
- CPU usage has dropped noticeably on the log server and is much more even between cores
- Tufin API calls were failing straight after upgrade and it turned out that each CMA needed at least one published change before API calls started to get accepted. Plus you should be running TOS 19-3 or 20-1 as a minumum. We are on 19-1 HF3 but it seems to work so far
Hope it helps someone! Will update as we go along.