upgrading to newer version on another server.

[santo]

I'm currently using Scalix 10.0.1 Community Edition on SUSE Linux 10.0.
Now (or as soon as it is released) I want to upgrade to Scalix 11 Community Edition on another machine
The reason for this is that I have bad experiences with upgrading SUSE Linux.
Therefore I want to install SUSE Linux 10.1 on a new machine (as 10.0 isn't supported anymore by Scalix 11) and upgrade/migrate my current scalix installation to that machine.

What's the best way to do this ?

Can I just copy some folders from Scalix 10.0.1 on SUSE 10.0 to the new SUSE 10.1 machine and then start the Scalix 11 installer there (which will then hopefully upgrade everything) ?

Or should I just start a new Scalix 11 installation on the new SUSE 10.1 machine and afterwards export all data from the old Scalix 10.0.1 server and import into the new Scalix 11 ?

[constant]

.
I agree with you about doing a new install instead of an upgrade, but if you have your drive partitioned, then do a new install and only format the / and /opt directories. That way all your home data, databases, web pages, etc., will stay intact.

I know this doesn't answer your exact question, but I thought it might be partly relevant.
.

[alfista]

I'm interested in doing the same thing here. Any additional tips?

[kanderson]

The main thing you'd need to do is transfer over /var/opt/scalix.

The gotchas here will be permissions & FQDN. They need to remain identical on both boxes.

Once you have copied over /var/opt/scalix, you should be able to install scalix, and the installer will detect the message store, and continue on smoothly.

Given the option, this is a great way to upgrade, because the old box is effectively a tested backup for you in case something goes south.

Kev.

[alfista]

Hi Kev,

Hopefully it is as easy as it sounds! I'll be giving this a try in the next couple of weeks.

Thanks a lot,

Jason

[msnead]

I have some similar questions:

We initially installed Scalix 10 for testing purposes to determine if it will meet the needs of our organization. Now that everyone is happy with it, I want to move the entire company over to Scalix (hopefully 11). Our original installation had a 50G /var partition. We have since purchased a SAN and now have 375G of space to allow for future growth. How can I move my existing message store over to the SAN? My thoughts are to copy /var/opt/scalix to the SAN but I'm worried about preserving permissions. Has anyone done this before that can provide instructions? Should I wait until after 11 is out?

Onto a different subject: Since the "window" of time to ask support installation questions was missed, can someone tell me the best way to set Scalix up for approx. 300 users? (No more than 50 premium and 250 standard users.) Basically, my real question is: do I need more than 1 server? (We have everyone on a single Sendmail box right now...does Scalix add enough overhead to warrant an additional box?) If so, what is the best way to split them up?

Thanks,

michael

[eltonmachado]

For copy issues you may use dump for the proper directories, the only worry is to see if the id of users mach. Dum is great for full backups.

[stevemurphy]

I performed a test upgrade using this approach tonight and it ended up being successful in the end, but took way too long to get there. I am moving from an existing Scalix 10.0.5 install on a FC4 server to a new server running FC5. I made sure the scalix username/group had the same id number on the new server as the old server and used scp to move /var/opt/scalix to the new server.

When I ran the installer, it gave me only the option to install and it did not seem to know how to handle the message store I copied from the old server. After the installer installed all of the components, it asked me if I wanted to remove and create a new mailstore and I answered no. Then the installer proceeded to ask me the usual questions for a new Scalix install (mailnode name, domain name, username format, sxadmin password, etc). After entering the same values as were used on the old server, the installer immediately fails. I think the issue is related to the mailnode name since the wrong mailnode name was in the error message in the installer's logfile, the tail of the logfile is below:

Code: Select all

2007-01-01 22:37:56,577 INFO Checking if any deprecated packages are installed.
2007-01-01 22:37:56,897 INFO Install end
2007-01-01 22:37:58,406 INFO Mailnode name derived from FQDN: mars
2007-01-01 22:38:04,342 INFO Mailnode is mars,blueslate
2007-01-01 22:38:28,986 INFO Domain is blueslate.net
2007-01-01 22:38:28,987 INFO Display name format is Firstname Lastname
2007-01-01 22:38:28,987 INFO Email addr format is Use display name format
2007-01-01 22:38:36,414 INFO Creating Scalix instance...(please wait -- this might take several minutes)
2007-01-01 22:38:36,538 CRITICAL Failed calling command '/opt/scalix/bin/ommakeom' with exit code 1 msg ['ommakeom: "mars" is already in the global configuration file']
2007-01-01 22:38:36,539 CRITICAL Operation failed: config-server function create_scalix_instance failed: Msg: Failed calling command '/opt/scalix/bin/ommakeom' with exit code 1 msg ['ommakeom: "mars" is already in the global configuration file']
2007-01-01 22:38:36,539 ERROR Operation failed: Aborting server configuration
2007-01-01 22:38:38,325 INFO Ending Scalix Installer - Msg: Failed to configure server


I tried uninstalling and reinstalling several times, experiencing the problem each time. I looking around a bit and noticed that /opt/scalix/bin/ommakeom is getting the mailnode value 'mars' from /etc/opt/scalix/instance.cfg (which did not exist on the old system, but does on the new system after the installer exited). I tried changing the mailnode name in instance.cfg from 'mars' to 'mars,blueslate' and was able to run ommakeom, but I was not able to start scalix afterwards:

Code: Select all

Starting Scalix services (mars,blueslate): omshowmn : [OM 6623] No mailnodes currently configured
                                                           [  OK  ]


At this point, I tried running the installer again to uninstall/reinstall scalix. The installer seemed to recognize the mailstore correctly this time. I was not asked for any of the mailnode information I was asked for before and the installer ran to completion without any errors. Afterward, I was able to log into webmail and sac without any issues.

Since this was only a dry run of the migration from the old FC4 server to the new FC5 server, I will have to do this again next weekend with a fresh copy of the Scalix 10.0.5 mailstore before decommisioning the old server. Is there any way to avoid the issues with the mailnode name 'mars,blueslate' that I ran into during this upgrade attempt? Ideally, I would like to be able to perform this upgrade with only the Scalix installer after I copy the /var/opt/scalix directory to the new server. Is there anything else that I need to copy in order for the installer to detect a Scalix 10.0.5 message store correctly?

Please let me know if there is any information that I can provide from either system that would be helpful.

Thanks,
Steve

[florian]

and used scp

You should not use SCP on the directory tree. SCP does not preserve file and directory permissions when copying across the network.

You might be able to fix this using the omcheck tool (see the manpage), however, most likely this will be a big permission mess.

What you want to do is to pack-up all the files using tar (which preserves permissions), copy over the tar archive and then unpack it on the other side.

Hope this helps, happy new year,
Florian.

[stevemurphy]

Hi Florian,

I tried using tar this time and the installer worked the first time. Thank you!

Steve

[mdraghici]

First, congratulations to the Scalix team for a great product and a Happy New Year to everyone.

I am trying a similar setup - moving to scalix 11 from scalix 10 on a new server and preserving the message store. My old hardware is running Open Suse 10 while the new hardware is running Open Suse 10.1

First I tried installing Scalix 11 on the new machine without bringing over the mesage store and everything went smoothly. Then, I uninstalled completely and followed these steps.

- created on the new machine group scalix and user scalix with the same UID and GUID as on the old machine
- copied the whole directory /var/opt/scalix from the old machine to the new machine using
a NFS share (I checked and the permissions were kept and everything went smoothly)
- when I start the installer, it sees the message store and runs through most of the instalation steps but fails when it tries to run the scalix server.

the error is:
ERROR: Operation failed. Starting Scalix Server failed: /opt/scalix/bin/omrc: (1) ['ERROR: Inconsistency found between version files in sys and lib', ' directories. This means that the patch levels of the Scalix shared', 'library and binary areas do not match. Scalix cannot be started until'. ' this is resolved.', '']


When I uninstall and start a blank installation everything works fine again so I am guessing the problem lies somewhere with the import of the message store.

Does anyone have any ideas?

Thanks in advance

[Valerion]

Sounds like the message store upgrade never happened (the installer runs ompatchom for this, but may do more).

I would suggest uninstalling Scalix and all its components, keeping the mailstore in place, then re-installing them.

[mdraghici]

Thanx for your help. I managed to move the mailstore and the upgrade went smoothly.

My problem was that when I exported the directory /var/opt/scalix in the NFS server I used root_squash option. This meant that I was not copying all the files from the scalix directory minus those that were owned by root. You need those as well so you need to use 'no_root_squash'

hope this helps somebody out there

best regards

~mircea