Install wont detect existing message store

[mito]

I'm not sure what's going on here...

I have a system that I setup on Fedora4. Instead of updating it to Fedora5 and then Scalix 11, I decided to wipe/reload it with CentOS 4.4, restore Scalix 10, and then upgrade to Scalix 11.

I have gone through a test-upgrade twice... once on a VM server, once on an old piece of hardware, both times I was eventually successful in my upgrade (the first time I had some problems with restoring Scalix 10, but I figured it out and had zero problems with my second test). So, since the second test went without problems, I proceeded on with the real thing. Thankfully though I kept a ghost-copy of my source setup, as I just finished reverting back to that ghost drive for now.

I have outlined my procedures in my own wiki (I planned on possibly adding it to the Scalix wiki after it was finished if it was good enough). BUT, this time through the upgrade, no matter what I did, the installer(s) would not detect that there was an existing message store!

I tried both the methods that worked for me the in my first two test-upgrades (note that I'm pretty sure the reason my "Method 1" didn't work the first time is that I didn't move the restored scalix dir, I copied it.... thus why I discovered the 2nd method) but neither worked. It would always just skip past the part where it would say that it found one, try to create a new one, and then give an error saying that the /var/opt/scalix dir exists but is not empty. I felt like saying "Yeah, no duh... use the existing files then!" at that point!

The only thing that I can think of is that when I tarred my existing message store, it either didn't tar the whole thing for some reason, or it got corrupted during the file transfers. Since I'm using the same hardware, I couldn't copy it right from one machine to the other, so I shutdown all of scalix ( service scalix stop ) and then tarred it all up ( tar -cvf BackupScalix.tar /var/opt/scalix ).

Now that I typed that, I guess I didn't stop scalix-tomcat before I tarred the files... could that be the problem? I don't think I stopped that the other two times I made the backup file for my test upgrades though?

Also to note:

I tried
* restoring my data, and then installing Scalix 10
* installing Scalix 10, stopping Scalix, scalix-tomcat, restore data, install scalix 10
* restore data, install scalix 11

Between each one I would uninstall all existing (leftover from the failed installs) setups via the installer, and I always deleted the previously restored data and then restored fresh. After playing with the scalix 11 installer a little, I realized that there were some leftover files in different places (/etc/opt/scalix had some files, /opt/scalix had some, etc) so I deleted any folder/file that had the word scalix in it that wasn't part of the installer, or my backup, and then I tried the first and third steps above again, and they still didn't work.

Anyone have any ideas as to why my message store wasn't being detected? It worked perfectly on both of my other two tests!

Thanks for the help!

[mito]

Anyone have any ideas as to why this is going on?

I haven't been able to test anything since I posted as it's still running the existing configuration (I reverted back to my backup copy until I figure this out).

One thing I'm going to try to do is copy the tarball of /var/opt/scalix that I tried to restore into one of my test systems and try to upgrade that, see if it's possibly something wrong with that tarball...

Other than that, anyone have any other ideas as to what could be going wrong?

Mito

[davidz]

I'm no expert for sure, but I am attempting this exact same procedure on my production server (currently FC4 and Scalix 10 going to CentOS 4.4 and Scalix 11), and I have done the same things. I installed fresh on an older machine copied my data over then installed 10 made sure it worked then upgraded to 11. One thing you should check that I came across in my research is making sure the permissions of the tared files is the same. I did this doing:

Code: Select all

tar -cpvf scalixdata.tar /var/opt/scalix/*

The switch 'p' maintains permissions in the tar file, so when you unpack them they are the same. That's the only thing I noticed in your post that caused me grief when I did the same thing.

[mito]

One thing you should check that I came across in my research is making sure the permissions of the tared files is the same. I did this doing:

Code: Select all

tar -cpvf scalixdata.tar /var/opt/scalix/*

The switch 'p' maintains permissions in the tar file, so when you unpack them they are the same. That's the only thing I noticed in your post that caused me grief when I did the same thing.

Well, that shouldn't have mattered since I didn't do that on my other two tries, that worked, but the archive that I had backed up didn't work on my test systems either. So, I waited until the next weekend (today) and tried the update again, and that time it worked.

Well, I'm having a different problem now, but that's for my next post :)

[ScalixSupport]

Hi!

Preserving permission is required, as there are dependencies of file permission. Another
thing you could have tried instead of directly using the command below for backup:

tar -cpvf scalixdata.tar /var/opt/scalix/*

you could have used:

Code: Select all

[root@subir-rhel4 ~]# cd /var/opt/
[root@subir-rhel4 opt]# tar --preserve --same-owner -cvf scalixdata.tar scalix/*

the above command will restore the owner, permission and order. When your new
machine is ready, you can untar the backup file, created a user "scalix" with same UID
and GID on the target machine, and try the installer, now you should be prompted for
existing mail store. If not, stop the installer, try installing from command line and use the
--debug switch along with the installer, like:

Code: Select all

./scalix-installer --debug

This option should give you an idea why the installer is missing the existing message
store.

Try the above steps. I am sure they would be helpful.

Thanks,
Subir