The server here had a tough time of it the last two days, sorry if you tried accessing the web server or incoming email during this time. It all started when I ran out of disk space on the root, so I finally set up /usr on a separate HD the way I'd been intending for some time. In my zeal for cleaning up, I accidentally whacked python and spent some time bringing it back by referring to my notes. Now lots of space on the root where my home directories live. Next, I tried installing Mozilla Firebird, since I was getting tired of the quirks of Opera 5 and never ran Netscape. It complained that my libstdc++ library was missing. A search revealed that the version it was looking for was available, but only for RedHat 7 and up (I was running 6.2). Well since I was attending to old server loose ends, I took out the installation CD for RH7.2 and decided to go for it.
I did take some precautions first, saving some config files and all the user directories in case of catastrophe. The upgrade, though slow (it's only a 133MHz machine) was pretty uneventful. The fun started when the system came back up _sans_ some important services:
- Apache complained about a bad line in its config file pointing to PHP4. I don't actually use PHP, but was looking forward to playing with it once I'd upgraded it from version 3 to 4. Commented this out for now.
- While I was sitting at the desk I noticed that the hard disk was making lots of noise. top revealed that sendmail was taking lots of CPU, and it didn't take long to find out that my 'security' account was getting bombarded with error messages from xinetd. It would only begin once an X session had begun. What was happening was that the new Nautilus windows manager was trying to connect to sgi_fam and getting denied over and over again, many times a second. It was almost impossible just to delete all the mail coming into security before new items would come in. Based on some more google searching, I ended up editing sgi_fam's entry for xinetd to disable it.
- Next, I found that sendmail was not accepting connections from the outside world. Not this again! But this time, the problem was that it had to have an entry in hosts.allow or else tcpwrappers would turn all connections away.
- My blog code had some problems because the upgrade seems to have put Python 1.5.2 on in place of 2.1.1 which I was using. A symlink took care of that.
- CGI scripts were not getting served up, even though I think the conf file was correct. So I downloaded and upgraded to the latest apache 1.3
Now everything seems to be hanging together okay. Some weekend away from the office. And when I look at my homepage with Firebird, I see how deeply ugly its representation can be when you're not looking at it with IE, Mozilla, or Opera (sorry anyone who had to see that also).