Very Computer

by **Jason M** » Wed, 19 Aug 1992 22:35:11

HP 9000/847, running HPUX 8.02.

Our machine crashes from time to time, and during reboot we virtually
always see messages of the form

/dev/rroot: INCORRECT BLOCK COUNT INODE = 12345 (8 MUST BE 0) (CORRECTED)

We assume this comes from fsck doing its job, but what exactly does
'8 must be 0' mean? Eight blocks, presumably, but why should there
be zero of them?

On a related note, is '/dev/rroot' a standard UNIX thing, or an HP
thing? lssf gives:

# lssf /dev/dsk/c3d0s10
disc3 lu 3 section 10 address 52.6.0 /dev/dsk/c3d0s10
# lssf /dev/root
disc3 lu 3 section 10 address 52.6.0 /dev/rroot
# lssf /dev/rroot
disc3 lu 3 section 10 address 52.6.0 /dev/rroot

It appears that /dev/root and /dev/rroot are special links to the
disk partition with the root file system. Why do both of these
exist? Could they ever be different? Who uses them?

Thanks in advance,

by **Jason M** » Wed, 19 Aug 1992 22:36:40

HP 9000/847, running HPUX 8.02.

Symptoms of a recent system hang:

The machine responded happily to ping packets, and telnet would connect,
but the login prompt would never appear for telnet, we couldn't log in
on the console, and ftp/rlogin/etc. did not work. NFS requests would
all time out.

We rebooted the machine and now all is fine, but I've never seen these
symptoms before. Has anyone else seen this happen?

Thanks in advance,

by **Barry Margol** » Thu, 20 Aug 1992 04:00:04

>The machine responded happily to ping packets, and telnet would connect,
>but the login prompt would never appear for telnet, we couldn't log in
>on the console, and ftp/rlogin/etc. did not work. NFS requests would
>all time out.

A common cause of this is filling the process table. Ping responses are
done by the kernel, so they don't require a new process. Inetd accepts
connections before trying to start up the server process; in the case of
telnet, this means that telnetd can't run.

Another possibility is using up swap space.

However, I'd expect there to be messages on the console with something like
"no more processes" or "no more memory". But if the console output is
going to a window, the problem could be afflicting the window system as
well, preventing it from displaying the message.

--
Barry Margolin
System Manager, Thinking Machines Corp.

by **Fred Appelm** » Thu, 20 Aug 1992 03:27:57

Quote:>HP 9000/847, running HPUX 8.02.

>Our machine crashes from time to time, and during reboot we virtually
>always see messages of the form

>/dev/rroot: INCORRECT BLOCK COUNT INODE = 12345 (8 MUST BE 0) (CORRECTED)

>We assume this comes from fsck doing its job, but what exactly does
>'8 must be 0' mean? Eight blocks, presumably, but why should there
>be zero of them?

I am just guessing here: I think what that means is that just before
crashing the filesystem allocated a inode and some blocks that belong
to that inode. Somehow the number of blocks is administrated, but the
block themselfs not. Make the INODE block count 0 is another
way to tell that the inode is freed.

I do remember that the error codes that are produces by fsck are in
the manual. There is a seperate section that describes fsck.

Quote:>It appears that /dev/root and /dev/rroot are special links to the
>disk partition with the root file system. Why do both of these
>exist? Could they ever be different? Who uses them?

/dev/root is the block device, /dev/rroot is the raw device. Block devices
are used if you access the disk via the filesystem. Raw devices bypass
the filesystem and should therefore only be used on unmounted discs.

Fred

by **Wayne Kro** » Thu, 20 Aug 1992 07:36:12

:
: HP 9000/847, running HPUX 8.02.
:
: Symptoms of a recent system hang:

Could you tell me what the system was doing prior to the hang? What
devices were connected and which ones were in use? What were your users
doing on the system? Any system administration activity going on (e.g.,
running a backup utility)?

A tip: If your system has the battery backup option installed,
powerfail recovery will sometimes jump start a hung system. To initiate
powerfail recovery pull the power cord out of the back of the system for
a few seconds and then replace it (do NOT toggle the power switch on the
front of the system).

by **Norbert Pap** » Thu, 20 Aug 1992 09:41:40

> Jason May writes:

> Posted: Tue, 18 Aug 1992 13:36:4

< Org. : Cambridge Technology Partners

Quote:> HP 9000/847, running HPUX 8.02.
> Symptoms of a recent system hang:
> The machine responded happily to ping packets, and telnet would connect,
> but the login prompt would never appear for telnet, we couldn't log in
> on the console, and ftp/rlogin/etc. did not work. NFS requests would
> all time out.

This sounds exactly like a problem I used to have. The ethernet driver
on my 720/730s stopped accepting packets greater than 80 bytes (ping works,
ftp doesn't!) after it received an ethernet packet greater than 2096 bytes.
Note that the maximum legal ethernet packet size is 1514 bytes. But
some manufacturers of ethernet devices don't seem to take that limitation
too seriously. HP has a patch for this (PHNE_0469 on the 700 series, may be
the same on the 800s).

Quote:> We rebooted the machine and now all is fine, but I've never seen these
> symptoms before. Has anyone else seen this happen?

Yup, a reboot fixes it (until the next killer packet). BTW, so will
resetting the ethernet driver.

> Thanks in advance,

Hope this helps.

--

Prince Rupert Grain Ltd.
900 - 1055 West Hastings St.
Vancouver, BC, Canada