Re: Nasty Bug In AIX NFS ?
Date: 1995/07/16
Message-ID: <3ubb4a$8bc_at_giga.bga.com>#1/1
Steve Entwistle <steve_at_rama.demon.co.uk> wrote:
>Are there any known bugs in the AIX NFS server or client that would lead to
>data corruption ?
>
>Today I was installing Oracle from an NFS mounted CDROM and it failed
>while performing a link step (Oracle Reports 2.0, CDE 1 Prod 3) complaining
>that a library (libsrwii.a) was not in XCOFF format. I copied the file via NFS
>from the CDROM into a temporary directory and then copied the file from the
>CDROM onto the machine the CDROM is attached to and transferred it to the
>machine I was working on using ftp (in binary mode!). I ran cmp on both
>copies of the file and it reported a difference about 680K into the file.
>
>The problem may be data dependent since I performed the test on another
>(larger) file and there were no problems, and I have also installed software
>from an NFS mounted CDROM in the past with no problems.
There have in the past been such problems. It's been a while I think since there has been one of these. They are extremely rare and typically occur under extremely heavy load. What is more common is a bad router somewhere in your network that is corrupting data. This has happenned alot in the past.
Can you recreate the problem at will? Can you recreate it to multiple different servers? Can you recreate it on a simplified network with nothing but 2 machines and a cable between them.
If you can convince yourself it's not the network, I'd be suspicious of the cdrom drive more than NFS.
If you're on a 4.1 MP machine I'd be more inclined to believe you've found a bug since the code is newer and more complex.
Running iptrace on either the server or client with ipreport -rsn will help you know what's going across the wire. If the client is sending bad data that would narrow it down some.
-- curt_at_tkg.com -- SCTrace for AIX is $199 from http://www.tkg.com -- TKG inc.Received on Sun Jul 16 1995 - 00:00:00 CEST