I have seen this problem on most architectures (I haven't seen it on the
RISC/6000 simply because I haven't had logredo errors).  Basically, the
problem is that with a clean shutdown, the bosserver doesn't believe
that it needs to start a salvager on the next boot.

On everything but the RISC/6000, the AFS fsck (vfsck) runs and if it
encounters a problem, it will make a /FORCESALVAGE file at the top-level
of the affected filesystem.  However, there is a problem with this
methodology because the rest of the Transarc software doesn't look for
this.  I have modified the fileserver source code to look for this as it
starts up and if it encounters it, it will clean up what it has started
and exit (the bosserver then sees this exit and starts the salvager,
which will clean it up).  I didn't add filesystem specific knowledge to
the bosserver intentionally, since that is only a process-management
program.

For the RISC/6000, I have not yet developed a solution because of how
the filesystem is handled (it relies heavily on the operating system).
I have supplied patches for the other platforms to Transarc about a year
ago, and we have been running with it on all our servers at MIT for
about a year.

-Richard

------- Forwarded transaction

[1856]  [EMAIL PROTECTED] (Mitch Collinsworth) Info-AFS_Redistribution 07/14/93 
20:07 (28 lines)
Subject: Re: AIX logredo process messes up /vicep? ??? 
To: "Frank Swasey" <[EMAIL PROTECTED]>
Cc: [EMAIL PROTECTED], [EMAIL PROTECTED]
In-Reply-To: Your message of "Wed, 14 Jul 93 10:02:32 EDT."
Date: Wed, 14 Jul 93 18:47:20 -0400
From: Mitch Collinsworth <[EMAIL PROTECTED]>


>I had a situation here today where AIX decided noone needed to log into
>one of my servers.  Since I was able to communicate with the bosserver
>on this machine, I issued a bos shutdown -wait (and waited for it to
>finish) prior to resetting the machine.  When the machine rebooted,
>logredo ran for all of the /vicepX partitions.  Since I had done a
>proper AFS shutdown, salvager was not run.  However most of the volumes
>on 8 of the 10 partitions were offline until I manually ran salvage on
>the machine.  Has anyone seen this problem or one like it before?

I guess I have seen one like it.  I don't know what logredo is and I haven't
used AIX systems for servers yet.  When I was using a VAX for the server
everytime it went down unexpectedly it would come up and run vfsck and then
come online.  I don't recall if salvager ran automatically or not, but several
volumes (same ones every time) would remain offline until manually salvaged.
Transarc's response was something to the affect that they knew about the
problem but hadn't fixed it.  I've recently moved the server to a DECstation
and a newer version of AFS and haven't seen it reccur (though it hasn't been
very long since the switchover).

-Mitch Collinsworth
 [EMAIL PROTECTED]
--[1856]--

------- End forwarded transaction

Reply via email to