A while ago we were experiencing an odd issue on one of our HP-UX nodes. Users who logged in via ssh were having extremely slow login times. We run a two node cluster and only one was affected so we weren’t quite sure what to look for. Oddly users who logged in via telnet did not seem to be affected. A few calls around and we were able to diagnose that the wtmps file was huge (didn’t record how large exactly). Every time a user logs in ssh parses the wtmps file to check for your last login. Obviously the bigger the file the longer it takes to log in.
The fix is to reduce the size of the file. You don’t want to delete this file as it will cause problems in the future. But here is the quick emergency fix:
cd /var/adm cp wtmps wtmps.bak cp /dev/null /var/adm/wtmps
For a more permanent fix you can create a custom crontab entry (what we did). But there seems to be a lot of people suggesting logrotate as a way to manage all of your system logs on HP-UX. You may wish to give that a look.