Good App - Bad Install / Upgrade Process

[bonoboslr]

I wonder why Zimbra continues to wrap everything around a shell script and not use something like an RPM repo especially for upgrades.

[dijichi2]

have you taken the time to look at how it works?

[bonoboslr]

Not fully, but I have undertook the painstaking process of upgrading. And should that process fail at one little hurdle - then - no upgrade for you. Re-Install is what you are going to have to do. RPM is so useful with scripts and triggers and macros. I was wondering why this technology has not been used to better effect than one massive encapsulating shell script.

[uxbod]

Platform independence ? What issues have you had ? Been through 10 upgrades and not one has ever had a issue.

[bonoboslr]

Ok - for starters http://files.zimbra.com/website/docs...x_to_5.0.x.pdf is nonsense. I need to upgrade to 5.0.10 from 5.0.9 and the first thing that script does is disable the cluster service. Erm.. fail. Because this umounts the volumes where the binaries reside.

[uxbod]

Ok ... Your a NE customer so have you raised any of these concerns directly through your sales manager ?

[bonoboslr]

Not sure that a sales manager is going to help me here - I will have to log a bug which will probably make it into version 6.5 if I am lucky, like my other bugs. But for anyone who is interested, here is the work around:

Hack your /etc/hosts to stop the install.sh moaning about it being 'ahem' wrong.
Edit your /etc/cluster.conf and modify the service name so that the install.sh does not just disable the server thus unmounting everything you need to perform the actual upgrade.
Disable the service manually and then remount the shared filesystems and bring up the clustered IP address.
THEN you can perform the upgrade. This is a bork as far as I am concerned. And with the number of users I have - 500,000+, it makes me extremely nervous.

[Klug]

Did you go to https://support.zimbra.com/ and read the second post on the main page?

The documentation is (once again 8(() not up-to-date.

[bonoboslr]

Thanks - I am checking now. But there is still an issue:

2) Cluster is to remain enabled with Zimbra services up
3) run './install.sh --cluster active'

How does 2) manage to run if 3) disables the service? 2) should say that you need to edit your /etc/cluster/cluster.conf and rename the service.

[dijichi2]

This has absolutely nothing to do with not using rpms for installs/upgrades, it's a bug/failure in cluster handling.

What about systems that don't have rpm and facilities that rpm provides? The install/upgrade code is complex enough as it is, splitting this into platform dependent versions will introduce bugs per platform and they will take longer to find/fix.