CentOS Upgrade brings down Server

CentOS 4.2 auto-upgrade via Yum shafts MySQL

[AlexK]

Yum is run each morning via cron. This morning it took the whole distribution from 4.1 to 4.2, upgrading virtually (if not every) distribution and installed quite a few new ones.

The second stage after yum upgrade is to upgrade yum itself (yum upgrade yum) but this choked with a "run out of mirrors" message (who knew it could be so vain, huh?) So many servers requesting so many terra-bytes and, clearly, not distributed to enough mirrors to cope.

The next message was more serious - mysql died, and the website with it. On attempted restart there was a plaintive complaint from BDB; it was getting old and it's memory was going:

051014 10:49:40 mysqld started
051014 10:49:40 [ERROR] bdb: malloc: Cannot allocate memory: 3435973880
051014 10:49:40 [ERROR] bdb: PANIC: Cannot allocate memory
051014 10:49:40 [ERROR] bdb: fatal region error detected; run recovery
051014 10:49:40 [ERROR] bdb: fatal region error detected; run recovery
051014 10:49:40 InnoDB: Started; log sequence number 0 43635
051014 10:49:40 [ERROR] Can't init databases
051014 10:49:40 [ERROR] Aborting

My site does not yet depend on BDB nor InnoDB (thank God), so putting skip-BDB and skip-InnoDB in my.cnf was enough to get mysql back up and running, and the site back up on the web.

I am stunned that CentOS would allow yum to make a point upgrade of this kind, but more concerned to know what on earth is shafting BDB/MySQL, since one or the other is essential to my site's next point upgrade.

All intelligence gratefully received.