A/UX concerns

[William Roberts;]

In <1991Feb21.202509.11608 at ni.umd.edu> steveg at ni.umd.edu (Steve Green) writes:

>Alan Mimms (alan at apple.com) writes:
>>Excuse me, Steve, but I want to find out if there is something strange
>>about your NFS usage.  We here (as you might expect) have several HUNDRED
>>A/UX machines, all happily using NFS to mount things from a Cray, several
>>VAXen, a Solbourne, several Suns, DECstations, R6000s, Motorola Unix boxes,
>>and a host of others (including, of course, lots of A/UX machines).
>>I haven't heard anyone complain that NFS is unreliable in A/UX 2.0 or
>>2.0.1.  Is it possible there's something unusual about your site or how
>>you're using NFS that causes problems?

>Not one bit possible.  Could it be that the problem is fixed in-house but
>not made available to customers??

We have problems as well. Whenever I ask "Are there any known or reported 
problems with NFS and A/UX?" I get no answer at all (not even a definite No).
Mots of the people on comp.unix.aux who actively port things have reported 
this problem (look in the archives). It is a subtle timing problem because it 
isn't 100% repeatable, and it only affects certain combinations. Our Sequent 
Balance 21000 can be persuaded to do it (Dynix V3.0.14), and we have narrowed 
down the problem to using ld to write an a.out file onto the remote file 
system: no-one complains but you often get holes full of zeros in your a.out 
file. A file such as the qdsamp.c demo is big enough to suffer regularly.

>>I'm NOT trying to debug your problem, although I will happily attempt it
>>if you contact me.  I AM trying to say that A/UX NFS is NOT FLAKEY.

>YES IT IS!!!!  A/UX NFS is very flakey.  I cannot reliably compile
>ANYTHING on an NFS disk.

What machine are you using for a fileserver? Name, rank and OS serial number 
required. Has anyone else seen this problem on something up-to-the minutes, 
like a Sun machine running SunOS 4.1.1 say?

I have recently reported a bug, shown up by NFS in particular, whereby a slow 
machine (A/UX on a Mac II) using a Kinetics EtherPort II card will hang up 
talking to a Sun4 server. If you lower the rsize to 1024 the problem goes 
away, so the bug is provoked by NFS if not actually hidden in the NFS code. 
Contact UK.DTS if you can't locate this report.

Someone mentioned 2.0b9 - we hit very, very, serious NFS bugs in 2.0b10, but 
these were fixed in the 2.0 release. BTW, I am still waiting for a response on 
what was changed between b10 and the release to solve this problem...
--

William Roberts                 ARPA: liam at cs.qmw.ac.uk
Queen Mary & Westfield College  UUCP: liam at qmw-cs.UUCP
Mile End Road                   AppleLink: UK0087
LONDON, E1 4NS, UK              Tel:  071-975 5250 (Fax: 081-980 6533)