Upgrade from 11.4 to 11.4.1 has problems

[bhuffman]

I performed an upgrade from 11.4 to 11.4.1 on a RHEL 4.0 server and the installer hung at the end and the process needed to be killed.

After that, it seems that some components either were not upgraded correctly or at least the version information wasn't changed.

Now when I log-in to SWA, it shows a version of 11.4.0.11344.

If I do an "about" under help, I see the following:
Version 11.4.0.11344
Platform Version 11.4.0.1922
Server Version 11.4.1.11855

However the RPMs on the server look right:
[xyz@xyz~]$ rpm -qa |grep scalix
scalix-postgres-11.4.1.2072-1
scalix-chardet-1.0.20071031-1.rhel4
scalix-sac-11.4.1.4516-1
scalix-platform-11.4.1.2072-1
scalix-sis-11.4.1.1040-1
scalix-iconv-extras-1.2-1.rhel4
scalix-server-11.4.1.11855-1.rhel4
scalix-res-11.4.1.4516-1
scalix-mobile-11.4.1.876-1
scalix-libical-0.27.20071008-1
scalix-swa-11.4.1.11567-1
scalix-tomcat-5.5.25-415
scalix-tomcat-connector-11.4.1.896-1.rhel4

Any insight would be appreciated.

Thanks,
Brian

[schmoe90]

Did you restart everything after the upgrade? That could be why the installer hung - it couldn't do the restart.

[bhuffman]

Yep - and did a reboot....

[seanyseansean]

I have the same problem.

The install failed on installing the scalix-server rpm. Everything reports 11.4.1, but the webmail is still 11.4. The imap spinning bug I reported in 11.4 however is fixed, which makes me think the core is updated.

I have to say each time I do an upgrade on Scalix my heart skips a beat as there are always problems, some related (I think) to have a failover heartbeat/drbd setup. I still have to create fake ifcfg files for our bonded interfaces too. As Scalix is (AFAIK) only supported on rpm distros, couldn't you distribute the updates via a repository somewhere?

Also, our reseller advised us to never do upgrades - according to them we should uninstall then reinstall. This is a bit unacceptable, really.

Sean

[Valerion]

There's a lot of configuration that needs to be done in the installer, and doing that via post-install scripts will be much harder.

I prefer if my clients doesn't uninstall and re-install myself, usually upgrades go smoothly.

[seanyseansean]

There's a lot of configuration that needs to be done in the installer, and doing that via post-install scripts will be much harder.

I prefer if my clients doesn't uninstall and re-install myself, usually upgrades go smoothly.

It does seem a bit strange though - other equally complex software is installed via repositories with no problems at all.

But this is a digression really. Does anyone know a clean way of forcing a reinstall of the webmail to 11.4.1?

[seanyseansean]

Actually unlike the OP, my platform version is correct (11.4.1.2072) as is the server version (11.4.1.11855), it's just the SWA version this is out of date (11.4.0.11344)

Also - in every upgrade i've done i've had an error relating to unlinking a sx user. Apologies for including the log here, but I was a bit quick on the rm command

[seanyseansean]

I had to uninstall in the end, upgraded first but that failed due to the sxqueryadmin user problem and the crash at the end of one of the rpm scripts. Deleted the user (had to reinstall just the server bit!), then it worked.

A couple of points though:

When the install / upgrade fails and exits, the services and listeners are all running, even when I delete the rpms! This means I then have to go and kill -9 anything Scalix related which takes ages.

I still have to create fake entries in /etc/sysconfig/network-scripts for our heartbeat controlled virtual IPs otherwise the network detection fails. This is crazy, there should be a warning surely but having to fake my config to get this working is a bit picky.

Performing an upgrade rewrites some of our config. Now some of this I understand (the graphics for example as pre 11.4.1 there wasn't theming support) but our smtp port?! And the server certificates?

What should have been 1/2 hour of work followed by a few hours testing on Saturday, turned into a full day of a dead server on Saturday, as well as most of today. Now some of this may be down to the fact we run a failover setup, and have 2 instances that can run on either or both, but it's sufficient enough for us to get our reseller to do upgrades from now on - when it shoud be a trivial exercise for a point release.

Good points - the new planning view and the fact the old imap 'killer email' bug seems to be fixed. I also discovered the Zimbra web client which works really well as a basic Scalix front end - it's also a lot faster than the official one, and I might recommend it as a default interface for people who don't have complicated calendar requirements.