This morning I found that mysqld had been down for about 2 hours, leaving all my sites offline. When I try to restart it, I find that the whole mysql is MISSING (eek!). After getting it back installed, I start digging in the logs and find that the culprit is rhn_check which took the liberty of adding users to /etc/passwd and making a bunch of installs. Apparently, it didn't like my own installation of mysql and thus deleted it!

Why is rhn_check automatically updating packages like this? It's the first time in almost 2 years that I have this problem. Fortunately I wasn't on vacation or anything... eek.

Another fun thing it did is 'fix' /usr/share/rhn and /etc/sysconfig/rhn so that a manual up2date is now not possible.

I am having the same problem. Up2date went and installed a bunch of new packages and wrecked my DNS settings.

I also had a bunch of user ids created as well.

Not good!!!

I had user ids created as well.

It was rather surprising to find that the reason I could not restart the mysql server was because the server binaries had been deleted.

I didn't have problems with mysqld. It restarted fine for me. Do you have a custom configuration / installation. I stick purely with the RH RPMS.

My BIND configuration was wrecked instead. Luckily, I do nightly backups and was able to restore fairly easily.

I would like to see SM's reponse to this. I spent a couple hours investigating the problem thinking that I had been rooted.

For some reason my mysql packages (4.1 installed later) were replaced with red hat's 3.5 packages and the server went bo bo . WHTF

What would you like them to say? This isn't a SM problem. This is a RH problem and one that is responsablity of the Administrator (that is YOU) to fix.

Apparently up2date upgraded the system to new release (Update 6) automatically. In the process it delete all packages it didn't like and installed it's own :shock:

I wasn't aware that up2date does changes unless it is set as a cron job etc but obviously it does. Now I gotta hunt down and find out how to turn it off.

this update screwed up my dns configuration as well. It took me a while to figure out what was going on and then for a few hours I was almost certain I had been hacked as several files including /etc/passwd had been modified.
Then i noticed that up2date did a massive update and I came here and noticed this thread and now my blood pressure has gone down significantly.

It screwed up your stuff because you did not have the packages set in in up2date config to skip. These packages should be updated manually only with up2date -uf packagename and then the proper post update operations done (i.e. on cPanel after updating caching-nameserver, you MUST run /scripts/rebuildnamedconf).

I did have php and mysql in the skip list, they still got upgraded (and my newer mysql packages removed)

Question: if I remove packages with e.g. "up2date -e httpd", will apache updates automatically be skipped then? The apache server that's running is one that I want to compile from source.

No, it will still update. Here's the skiplist I use for a cPanel system:

your mod_ssl statement at the end is not needed, its covered by the mod_* one

Ooops, good call

woah! Same thing happened to me on the 28th. I didn't notice it until tonight because all of my sites were still up. MANY MANY packages installed. WTF happened?

BTW, mine didn't just update RPM's I had installed, it installed NEW packages. Tons of them!

The cool thing is that it first deleted packages it didn't like and then decided it was unable to install the upgrade. Nice, nice. I disabled all automatic updates now. I really thought rhn_check just did a check and not an update. It's less of a burden if up2date messes things up because at least I can fix things immediately.

Serhat,
How did you disable automatic updates?

Currently, I've disabled rhnsd from launching at boot. Will that do it?

That's what I did too... plus killing the rhnsd process. I think that'll do it.

Thank you, Serhat. Hopefully this won't happen again

I seriously think every RPM released by RedHat was installed on my server. It's an absolute mess

I took a peek at my server, everything thing seems to be going well.

I've got RHEL3 too, and up2date lists a half dozen minor updates I haven't bothered to install. My installation of MySQL 4.0.23, which I installed via MySQL's own RPM packages, is thus far untouched and still running fine. I just restarted it to make sure.

I'm just going to not touch up2date and see what happens. up2date should not run automatically on my box, as it never has before. If the ServerMatrix techs log in and run it and things go wrong, then I can justifiably demand that they fix any problems that occure since the server was working fine before they touched it. If nothing goes wrong, fine. So far up2date seems to be working fine for me.