Distro upgrade to sarge/stable on Debian VPS

[valor]

Hi All!

Does anyone already did a apt-get distro-upgrade on rimu VPS?

I'm scared libc6 udpate may cause problems with uml software... any thoughts?

regards

Guzmán

[carl]

Many times. :) It seems to go quite smoothly.

All I do is change 'stable' to 'sarge' in your /etc/apt/sources.list, then do:

apt-get update
apt-get -y dist-upgrade

If you need a hand with it, put in a support ticket and we'll help out.

[valor]

Hi all!

I don't know how you did it, but for me it was a real dissaster.

Yesterday I choosed the night to do the upgrade due to less users using my server services, after upgrade end suddenly everything was working except mysql.
I did the upgrade following line by line the Debian 3.1 Releasenotes.

Mysql had a segfault and after googling without find anyone with the same problem. Then I send a support ticket because without mysql my hosting doesn't worth a penny.

rimu staff sshed to it and after look around he offer me to do the upgrade from me restoring a backup from Jun 19 (Four days ago), he did the same upgrade I did (except he removed in the way apache, courier-daemon. squirrelmail and lot of perl libs packages), after he finish the upgrade he had the same problem.
This time he hit the problem, and it was glibc with tls support, tls support of glibc is broken on rimu VPS, he moved /lib/tls to /lib/tls.broken and mysqld started ok.

Operation cost: Upgrade to sarge got all my domains and users back to Jun 19, meaning that all e-mails they received in the meantime were lost, (this indeed was my fault, because rimu staff leave a /backup dir for me to get everything back as it was before the backup restore, but because it was 6:00 AM i felt sleep, suddenly I wake up now and because qmail is running there is nothing I can do to fix this problem... today will be a long complains day.

So, solution?
NEVER upgrade from woody to sarge without ask first rimu staff to do a full backup of your system first so if you have a problem later you don't need to go four days back to have a backup.
If you screw it, try to ask for a full backup of your screwed system, I wasn't smart enough to ask for it. If I had a backup of the screwed system, after rimu staff find the same problem, they would restore the screwed back and not apply the fix to the June 19 restore, which means not loose e-mails, web stats, etc.

Kind Regards

valor

[valor]

Hi All!

Lucky me, when rimu staff did the upgrade, he accept the new /etc/passwd and group file which removed the user and group vpopmail & vchkpw.
So, e-mail delivery was broken since sarge update, so I can restore e-mail from last point, anyway no one received an e-mail today because permission denied to vpopmail user.

If anyoen ever find on this situation, the procedure to restore it's easy.

Awstats restore:
Find your db directory where awstats stores information. Move it away and move the one on the backup, then for every virtual hosting restore the apache logs file to the last backup in the same way (mv original old; mv backup original)
Also if you added a domain after last backup, restore /etc/awstats files in the same way.
If you use plugins GeoIp from MaxMind or hotinfo you must install from source again the maxmin geoip and apt-get install libnet-xwhois-perl because of upgrade of perl your old xwhois will be removed.

Vpopmail Restore:
Easier than awstats: mv your old vpopmail to a new dir just in case and move the backup vpopmail dir to where your vpopmail dir was (usually /home/vpopmail), check that everything belongs to vpopmail user and done!

regards

valor