I had a somewhat bad experience yesterday with the upgrade feature when trying to upgrade a NetWorker server from version 220.127.116.11 to 8.2.2. I did the recommended intial steps before an upgrade (nsrim -x, nsrck -m, nsrck -L6 etc etc) and when trying to update, something went wrong and setup dialogue told me it changes would be rolled back. Fine, s**t happens. But when I tried to restart the NetWorker server I noticed that all NetWorker binaries were gone and that the NetWorker server was completely uninstalled from the system.
Anyone else experienced this? This was of course on a Windows OS.
Of course it is Windows OS - it only can happen there Most likely some mess in registry caused this to fail. Actually, I dare to say that most likely same reason stands behind update and rollback failure in my experience. Now, which one is it, I have no clue. Did you do update or did you uninstall and then install new version? Sounds like first one (I prefer second one). Did you use standard paths or relocation? Did you try to install it manually (w/o update option)? You can also run installer in debug mode which should create text log file and it might help to see where it breaks.
I did an update. Which of course is the official way of upgrading. However maybe not the most stable. As you say, an uninstall of the old version and then an new install of the new version is the best method. Standard paths was used. When I noticed that the rollback action also had uninstalled the old version I tried to install the new version manually but with no luck.
I will try to install the new version again and use the debug mode. Also Im now aware of alot of Windows patches/updates are waiting to be installed on the server. That que could also be a potential problem in this case.
I have also downloaded the latest cumalative update of NW. Hopefully that will help too....maybe....
We did our update last week (W2KR8) and it had absolute no issue.
I was surprised that you said "when I restarted the NW server ..." because this is what NW does automatically.
But I do not have a clue what you have/your system has done to make this happen.
In the meantime I update all NW clients via client push and I was positively surprised to see that the success rate now was about 95%. Obviously EMC has implemented some major efforts here.
3 obstacles remain:
- nsrcpd does not terminate automatically.
- the "client push client jobs" remain actively in the jobs jb and you must manually kill them later.
These can be hundreds/thousands.
- A restart/reconnect of the software administration GUI is not possible without deleting the cpdb and restart from
scratch with a new repository and client inventory.
My personal experience.
Just a small update. Did a second try yesterday of updating NetWorker server from 18.104.22.168 to 8.2.2. No luck, neither with unistalling old version first and then installing the new version or updating. It seems from the event viewer logs though that there was some problem with the vc redist. I will try to investigate further.
Decided to try and upgrade to version 8.2.1 instead of 8.2.2 and that worked without any problems at all. So maybe the vcredist binary supplied with the 8.2.2 biniaries is a bit quirky.
Could be that vcredist (which is MS package btw) that comes with SP2 is newer and doesn't like something on the system. I guess if I wanted to play with it, I would try to see which version they used in SP1 and SP2 and then try to install independently version which is claimed to be part of SP2 by downloading package from MS.
Im gonna see if I can recreate the problem in a POC environment and do just that. But not this side of 2015.