Upgrade Scalix 9.4 to Scalix 11.0.1 - the right solution?

[NSC]

Hi,

i have to upgrade from scalix 9.4 on SLES9 to scalix 11.01 on SLES10. Basicly this should be an easy way: upgrade from 9.4 -> 10.0 -> 11.0.1, then unmount storage, mount it on the new server, install scalix 11.0.1 and run omscan - everyone should be happy, but it isn't so easy
* our scalix-store is near 420GB and has more than 7 million files
* omscan runs 17 hours
* our database is a little bit corrupt - a few lvm-snapshot-crashes made us a gift of 48 Missing Child and 8 Missing user directories
* to repair the database i have to wait until omscan has finished and then use omcontain to fix the problems
so, the whole update-process with backup, update and fixing the problems will cause a downtime of minimum 30 hours and the result is not a really clean database - that is not really my favorit solution!

in my opinion, to use sxmboxexp and sxmboximp should be the prefered solution and if the export and import is scripted that more than one userexports/imports run at the same time the speed is more than good - we are faster than waiting for omscan has finished and at the end we have a new and clean database. updateprocess was made in the order of the following steps:
* upgrade from 9.4 -> 10.0 -> 11.0.1
* sxmboxexp
* install scalix 11.0.1 on sles10
* ldapsync with option 2 (yes, the cn and the omGlobalUniqueId are like those on scalix9.4 - it comes from openldap!!!!)
* sxmboximp
* omtidyu
* create mime-cache with omtidyu
* starting scalix and all seems to work perfekt, but there are two problems that makes this solution impossible for us:
* an already created outlook-sync breaks with an error
* the rights on shared folders seems to be backuped and restored too, but it only seems like that - the acl shows that the user with the cn xxxx is granted, but user xxxx has no rights!

a long description of my problem and two short questions about:
1. is my 2nd solution really the golden egg or am I wrong - if I'm wrong are there any suggestions?
2. what could cause the two problems after sxmobximp?

thanks in advanced to all experts in this forum for your help - im glad for any suggestions or tips!
thank you
thomas

[kanderson]

Personally, I'd fix the mailstore data issue in 9.4, and then upgrade to 10 check the mailstore again and then upgrade to 11.0.2 (will be out soon, and you'll need time to fix 9.4 anyway).

If you're worried about issues, you could call Scalix Support, and open an incident, but this doesn't seem insurmountable.

As you've clearly used the tools before, I'll just advise that you run omscan repeatedly until it comes back clean. Then start the various upgrades.

omcpoutu will lost LOTS of information. Since the information isn't there, sxmboximp can't pump it back into the "new" server. You'll lose all rules, all delegation, all ACL information, etc. That problem will be quite public while it's being fixed. If you fix the mailstore, there might be some slowness for the next week while omscan is running (You'll need to run it at LEAST twice, but as many times as it takes until it comes back clean, remember) but that will allow you to retain all of the granular configuration, which is what generally drives people crazy when it's lost.

Also, at this point, I'd recommend SLES users wait at v10 for a while, until this bug is resolved. http://bugzilla.scalix.com/show_bug.cgi?id=14709 I don't think SLES10 is affected, but it's a nasty bug, and you REALLY don't want to see it. v10 is a fantastic release, and a significant upgrade from 9. My advice would be to wait there until this is resolved.

Kev.

[NSC]

hello kanderson,

firstly thanks for your answer!

Personally, I'd fix the mailstore data issue in 9.4, and then upgrade to 10 check the mailstore again and then upgrade to 11.0.2 (will be out soon, and you'll need time to fix 9.4 anyway).

This is the recommended solution, I know, but as I told, omscan runs 17 hours and it doesn't fix the corrupted database and so I've to correct it manually with omcontain - a horrible think in my mind!

omcpoutu will lost LOTS of information.

I know this, but sxmboxexp should not, but therefor I've to upgrade to scalix11!

I have a testinstallation of scalix11.0.1 on sles10 since 3 weeks and tested a lot of things to get the best solution how to migrate and there were three points that changed my opinion (firstly, I had the same as you):
* sxmboxexp and then sxmboximp results a clean db - just only two usersboxes of our 2300 users are dead and could not be exported - collateral damage backup 2 of 2300 users into a outlook.pst and rebuild their mailboxes is ok for me if I get a really clean db!
* i don't have to repair the database manually with omcontain!
* downtime, as I already told - yes, if I repair all in 9.4 the server could get online after upgrade and then let run the omscan in background, but how can I get sure that there are no problems left? if there is a problem with a mailbox indexserver and mimecache-creation stops at this point - yes, they don't ignore the corrupt mailbox and jumps to the next, they stop!

now you maybe understand why i would prefer the second scenario? (but if I won't get a solution for my acl- and sync-problems there is no other way as to fall back to your suggestion!)

I don't think SLES10 is affected

as I told, we have a testsystem with sles10 and it's not affected!


once again thanks for your answer,
thomas

[kanderson]

At this point, I'd recommend opening a support incident with Scalix to resolve the mailstore issues. That should be fixed before you upgrade. Omscan can run while the server is online, so I don't think there's going to be much of an impact on the end users beyond performance somewhat.

Getting to 11 with a corrupt mailstore won't make things easier for you. My (and I assume your) worst nightmare would be running sxmboxexp and receiving an error that prevents you from backing up your users as planned, as well as seeing that corruption mean a user is unable to log in at all.

My advice would be to have Scalix Tech Support fix the mailstore for you, and then you can go forward with the upgrade from there. It's not expensive compared to even a day of grief with your mailserver. Note that I don't think I've ever recommended calling Scalix support to anyone on here before. You've got a big server, and a REALLY big likelihood of serious problems. At this point, you can have it fixed easily and save yourself a weekend worth of work, as well as user frustration. Take that route. You won't regret it.

You can run omscan -Aavfx yourself to try before calling, it might work fine, but you're asking for problems to upgrade knowing you have a corrupted mailstore.

Kev.

[NSC]

Yes, you're right - I'm going to consider the Scalix support!

About to go to 11 with a corrupt mailstore:
I've tested my prefered solution as described, based on a full backup of the running system and therefore with all users and (near) the whole datastore - after finishing upgrade the omscan does not report any error, the users can login normally ... - as i told in my first post, there are only two things that are not working, but they are substantial!

once again thanks
thomas

[kanderson]

You can reach them at support@sclix.com (Correct the spelling, obviously).

Kev.

[SirAdam]

Personally, I'd fix the mailstore data issue in 9.4, and then upgrade to 10 check the mailstore again and then upgrade to 11.0.2 (will be out soon, and you'll need time to fix 9.4 anyway).

What it soon, can you give a timeframe when it will be available?

Regards

Michael

[kanderson]

This week assuming QA gives the green light.

Kev.