From: John Ladd (johnladd68@googlemail.com)
Date: Tue Apr 01 2008 - 09:08:46 EST
Hi there,
I ran across one of these 'disk full' messages on a UFS partition that had
free inodes and free space as reported by 'df'. Found a folder that had lots
of small files (around 450K of them), and any try to create a new file would
report a 'file system full'. Here the related error messages on syslog:
Mar 31 17:33:31 hostname ufs: [ID 845546 kern.notice] NOTICE: alloc:
/data/applcsf: file system full
Mar 31 17:33:35 hostname ufs: [ID 213553 kern.notice] NOTICE: realloccg
/data/applcsf: file system full
Mar 31 17:33:47 hostname last message repeated 2 times
After checking the usual things (as I said, df -k, df -oi) and not seeing
anything weird, I tarred some older files (that is, on another partition),
and everything came back to normal.
The folder is used to write some checkpoint/request/out files (this is part
of an Oracle application server setup), but there are about 20K new files
each day, and the partition is 'just' 20GB. Space reported by 'df' does not
seem to be an issue. When this error happened, capacity was reported to be
about 84%, and percentage of used inodes (%iused) around 54%.
Technical support asked things we had already checked and suggested to move
onto a zfs filesystem, although it's pretty weird there's no easy
explanation why this is happening. For what's worth, fragmentation on this
partition is 10% at the moment.
Anybody seen this before? I could not find any limitations on UFS filesystem
as per design (other than a 32K-subfolder for a given folder). Or shall I
just push it back to the end user and ask them to implement a cleaner policy
for archival/removal of old files?
Cheers,
John
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:42:53 EDT