Unsolved
1 Rookie
•
64 Posts
1
1770
Uninstalling Patch KGA_RUP_2019-05_249622 on OneFS 8.1.2.0 Failing to Complete on Simulator
Our cluster is on 8.1.2.0 and I use the Simulator of the same version to model our patch installs. Although the kernel patch KGA_RUP_2019-05_249622 installed normally on the simulator (rebooting each of the three nodes in turn), when I go to uninstall that patch it completes the uninstall on Node 1 and then just hangs in the "Uninstalling" state for that patch. The only way I can get it to progress is to clear the patch using the "scary" command that tells you not to proceed unless directed by an Isilon engineer, and reapply the uninstall command. Then it will do Node 2 and hang. Repeat everything for Node 3 to finally get it to go away. I even completely tore down the simulator nodes and redeployed from the .ova for a fresh start, but the results are the same on my second run. No amount of rebooting the nodes gets the patch to continue the uninstall.
I can't be certain this will happen on our production cluster, but I'm concerned because the Userspace patch UGA_RUP_2019-05_249620 installed and uninstalled normally on the simulator (although that patch doesn't force a reboot) and I was able to deploy the Aug version of that patch normally.
The simulator itself is unsupported but I'm going to try opening a ticket framed in these terms and see if I can get some help. Has anyone else experienced similar issues with a patch refusing to completely uninstall? This is our first go with uninstalling the kernel patch and installing the next version since our recent 8.1.2.0 upgrade (from 8.0.0.6) a few months ago. At that time, the kernel patch 249622 installed completely normally on the cluster (rebooting each in turn) just like it did on the simulator. Hard to think I won't have an issue once I attempt to uninstall the patch in production.
dle3
1 Message
0
October 31st, 2019 12:00
We experience the same issue on uninstall UGA-RUP_2019-08_254383 on simulator 8.1.2.0. The uninstall just hangs.
Ryan_CSULB
1 Rookie
1 Rookie
•
64 Posts
0
November 15th, 2019 09:00
EMC Support was gracious enough to actually attempt to duplicate my issue, which they couldn't do. It didn't hang for them, and worked as it should. I have to take them at their word. However, it still doesn't work for me and to be honest, something is really odd about the simulator setup itself. I can't put my finger on it, but it just doesn't seem to setup the same as the old 8.0.0.6 simulator I used to use... I can't get all the nodes to talk to each other reliably using a simulated infiniband switch for internal communication. Worked perfectly before, not so much now. That may be the heart of the issue for us and maybe we're doing something wrong (two of us here separately spent way to much time trying to get it work and gave up).
In any case, happily the update worked as expected in Production and went smoothly.
tenortim
36 Posts
0
November 15th, 2019 15:00
It's a bit of a SWAG since I don't have enough data here to give you a good answer, but what I think may be happening is that you're seeing a timeout due to the very different performance characteristics of the simulator.
One thing I would strongly suggest doing on the newer releases of the simulator is increasing the amount of memory allocated to each virtual node. Even going from 2GB to 2.5GB help, but if you can spare 4GB or better yet 6GB, it really helps. I recently opened an internal bug requesting that we change the default in the simulator because 2GB is just not quite enough any more.