[comp.unix.questions] i-nodes disappearing in SysVr3 filesystem

paul@dy4.uucp (Paul Burry) (04/27/89)

About a year ago, a problem was mentioned in this group where a bug in
the System V filesystem was causing either i-nodes or blocks to disappear.

Well, I believe I am now experiencing this very problem.  When running
inews, the filesystem becomes corrupted and the number of free i-nodes
suddenly drops from ~2300 to ~1.  Only performing a fsck on the filesystem
restores the i-node count.

I remember a discussion in this group about a year ago where someone
described a similar problem with the filesystem.  My question is
what can we do to avoid this bug ?  I know that the filesystem should
probably be remade with more i-nodes, but will this eliminate the problem
or merely delay it ?

Our configuration is:
	- Motorola Delta Series M8000
	- System Vr3v3 Unix

/         (/dev/dsk/m323_0s0):     5258 blocks     2120 i-nodes
                   total:    25272 blocks    3152 i-nodes
/usr      (/dev/usr        ):    11600 blocks     8851 i-nodes
                   total:   103680 blocks   12960 i-nodes
/usr1     (/dev/usr1       ):    93710 blocks     2375 i-nodes
                   total:   152928 blocks   10000 i-nodes

So far, the problem has only occurred on the /usr1 filesystem.

Any help would be appreciated,
	Paul Burry
-- 
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Paul Burry			
UUCP: ...!cognos!dy4!paul	POST:	Dy4 Systems Inc., 21 Fitzgerald Road,
PHONE:	(613)-596-9911			Nepean, Ontario, Canada K2H 9G1

bill@twwells.uucp (T. William Wells) (05/01/89)

In article <131@dy4.UUCP> paul@dy4.UUCP (Paul Burry) writes:
: About a year ago, a problem was mentioned in this group where a bug in
: the System V filesystem was causing either i-nodes or blocks to disappear.

Inodes.

: Well, I believe I am now experiencing this very problem.  When running
: inews, the filesystem becomes corrupted and the number of free i-nodes
: suddenly drops from ~2300 to ~1.  Only performing a fsck on the filesystem
: restores the i-node count.

That sounds like the inode bug all right.

: Any help would be appreciated,

Fsck'ing your system frequently will cut down on the failures but will
not eliminate them. I've had an inode crunch within minutes of doing
a fsck.

If you have source code, or are handy with a disassembler, it is
fairly easy to patch the code. I can give you hints if you need them.

Of course, the right solution is to hassle your vendor. If he tells
you that this is hard to fix, tell him baloney. The bug is in
(s5)?ialloc and is trivial to fix.

---
Bill                            { uunet | novavax } !twwells!bill

cgh018@tijc02.UUCP (Calvin Hayden ) (05/03/89)

> About a year ago, a problem was mentioned in this group where a bug in
> the System V filesystem was causing either i-nodes or blocks to disappear.
> 
> Well, I believe I am now experiencing this very problem.  When running
> inews, the filesystem becomes corrupted and the number of free i-nodes
> suddenly drops from ~2300 to ~1.  Only performing a fsck on the filesystem
> restores the i-node count.
> 
> I remember a discussion in this group about a year ago where someone
> described a similar problem with the filesystem.  My question is
> what can we do to avoid this bug ?  I know that the filesystem should
> probably be remade with more i-nodes, but will this eliminate the problem
> or merely delay it ?
> 	Paul Burry

I too am experiencing problems of this sort.  We are running AT&T Sys V
r2 v2 on a Vax 8600.  Ocassionally the /usr/spool/news file system 
shows as having 1 inode.  This is usually overnight that this occurs.
A script executes a df around 10pm, and then we get our newsfeed from
11pm to 7am.  The df show ok, but when I check it in the morning, the
free inodes has often gone from 2000 or so to 1.  FSCK reveals that I
have a bad free inode count in sblock, and fixes it, giving back several
thousand free inodes.  I have physically moved the file system, so I know
that the sblock isn't located on a bad disk.  

Example:  Last night 10:30pm:
   df reveals  /usr/spool/news (/dev/dsk/0s3):  47662 blocks  4065 inodes

          this morning 8am:
   df reveals  /usr/spool/news (/dev/dsk/0s3):  43121 blocks     1 inodes

	  after unmounting and fsck'ing it, and remounting:
   df reveals  /usr/spool/news (/dev/dsk/0s3):  43121 blocks   4022 inodes
 
I could increase the # of inodes for the file system, but will this
solve it.  It has happened several times in the past 2-3 months.  More 
of an aggravation than a real problem (as far as fixing it goes).

Same questions as Paul... Any suggestions?  Thanks in advance!!

++++++++++++++++++++++++++++++++++
+Calvin
+uucp:  ...mcnc!rti!tijc02!cgh018
++++++++++++++++++++++++++++++++++

"What is the airspeed of a laden swallow?"  Old man from scene 20something
                                            in MP's HOLY GRAIL