pha@cs.rit.edu (Paul H Allen) (07/24/90)
I maintain three 3B2/522s and one 3B2/600, each running SysV Rel 3.2.1 V3. The 3B2/522s advertise their user partitions which are mounted by the 3B2/600 where the faculty and students logon to do their work. All rfs is supported by Wollongong's TCP/IP WIN/3B Release 3.0.1. A while back the root partition on the 3B2/600 ran out of inodes. I checked the obvious, i.e., the tmp directory, all the mount points and found nothing. We have a nice message setup to relay info to the students that resided on the root partition. It has a small database that consumed a lot of inodes. So I moved it to /usr/spool and gained back over 400 inodes. This action gave me an immediate solution to my problem so I didn't both looking any further. So last week, the root partition ran out of inodes again. All I could think of was the infamous loss of inodes that commonly plagues news partitions, at least it did mine. So I put the 3B2/600 in single user and did a fsck on root. I couldn't believe my eyes, all I saw was `LINK COUNT TABLE OVERFLOW (CONTINUE?)' scroll by during Phase 1 and many requests to CLEAR files during Phase 4. It took between 80 to 100 fscks before the file system became clean. I booted off the Operating System Utilities tape to obtain a standalone shell just so the machine wouldn't reboot after each fsck. By the way, does anybody know how to boot to single user from firmware? I also checked the root file systems on the 3B2/522s and found them to be corrupted to varying degrees. Quite a while ago there was a small discussion regarding how it wasn't a good idea to mount file systems without doing a fsck just because they were unmounted cleanly. Would it be OK if I inserted a mandatory fsck entry in the /etc/init.d/MOUNTFSYS file just before the mountall line? Also the 3B2/600 has been crashing about every 2 or 3 days with a PANIC: Kernel MMU fault (F_ACCESS). Could this be related to the root file system problem? Diagnostics reveals nothing wrong with the hardware. My next step will be to reload the operating system (uugh!). Have you done a fsck on your root file system lately? After fsck finishes, there is a summary of x files checked using y blocks and leaving z blocks free. The first fsck listed about 3100 files and the last one listed less than 1000 files. Unfortunately, the number of files checked is reduced by 19 for each fsck done. Any guidance or recommendation would be appreciated. Paul Allen (716) 475-5254 pha@cs.rit.edu ...!rochester!rit!pha pha1775@ritvax (BITNET)
jrallen@devildog.att.com (Jon Allen) (07/26/90)
In article <1777@cs.rit.edu> pha@cs.rit.edu (Paul H Allen) writes: . . . >Also the 3B2/600 has been crashing about every 2 or 3 days with a >PANIC: Kernel MMU fault (F_ACCESS). Could this be related to the root This is actually a bug in TCP/IP 3.0.1. You should be able to call your AT&T support line and request a fix to the TCP/IP F_ACCESS MMU fault problem. Many of my machines crashed several times a week for 6 months before I got a fix in April. Since April, I have had no problems. -Jon