Greetings -
--On Monday, April 17, 2000 12:31 am +0200 Andy Polyakov
<[EMAIL PROTECTED]> wrote:
>> On our single-processor systems the ssh-2.1.0 server seems to work fine.
>>
>> However on our 6-processor Solaris 2.6 box it doesn't.
>
> Any progress on this? I can't reproduce it on 2 CPU 2.6 box, nor 4 CPU
> 2.5.1 box...
I think Sami is looking at it (the message disappeared into one of his
procmail filters for a while before re-emerging).
I'm definitely seeing the problem on our 6-CPU Solaris 2.6 system...
When I tried it again today I thought for a moment it had cured itself.
However it turned out that therre were lots of CPU-hungry jobs running on
the machine, making it pretty much effectively the equivalent of a single
CPU box:
Out of 10 attempts to connect 9 succeeded (90%) and 1 hung (10%).
(Previously, when the system was virtually idle I was seeing a connection
failure rate of nigh-on 100%.)
I then re-niced down the CPU-hungry jobs to give sshd more of a chance at
grabbing a spare CPU. This gave:
Out of 20 attempts to connect 14 succeeded (70%) and 6 hung (30%).
I then tried using "pbind" to tie the sshd to a single CPU, and got:
Out of 10 attempts to connect all 10 succeeded (100%).
I'm pretty convinced there's a race-condition lurking, and am hoping and
praying Sami (or someone) can spot it. (Do let me know if there's any mods
you want me to try. However note that I'm on holiday 19th April - 30th
April inclusive.)
In the meantime I'm going work around the problem by altering the sshd
startup script to start it up, sleep for a couple of seconds, then "pbind"
it to a single CPU.
Cheers,
Mike B-)
--
The Computing Service, University of York, Heslington, York Yo10 5DD, UK
Tel:+44-1904-433811 FAX:+44-1904-433740
Web: http://www-users.york.ac.uk/~pmb1/
* Unsolicited commercial e-mail is NOT welcome at this e-mail address. *
