Today, one of our BCRC computers ran out of space on the ssd. The
symptoms were a gray screen with a pointer that could move, but nothing
else was visible. At some point, after rebooting, I saw a dialog box
saying that the startup disk was full, but when I tried to log in, I got
the same result.
I rebooted using the recovery permission and opened a terminal. I didn't
see any obvious problem. I removed a few cached user home directories
from /Users/OneWeekBackup so I could reboot the system and have a more
functional environment. (The rescue shell doesn't have, for example,
"du", which is useful in this circumstance).
After rebooting, I began looking for where the big files were and I
found /private/var/folders. It was defined in the negative transcript as
a directory that radmind should leave alone but, after doing some
searching, I found that its purpose seems to be userland cache and temp
files. And that these files are not deleted even when users are deleted
so, in our context, they can grow almost without limit. I removed it
from the negative transcript and referenced it in the base transcript --
this means that the directory will exist, but the contents will be
removed. After running radmind, it had recovered ~100GB.
Currently OneWeekBackup is set to 15 days. If we start to run into space
problems, we can reduce it to 8 days. But it seems premature to do so now.
I've made this change on snapper and tested it in the BCRC. But I wanted
to mention it to you guys before I sync it to wahoo where the change
will affect your machines.
Cheers!
--
Steven D. BREWER <sbrewer_at_bio.umass.edu>
http://www.bio.umass.edu/biology/about/directories/faculty/steven-d-brewer
Senior Lecturer II; Director, Biology Computer Resource Center
Titolo sen mono--sensignifa sono.
Received on Tue Oct 21 2014 - 14:04:52 EDT