#net rendering question
5 messages in this thread
I have a basic rendering farm started and it's working correctly except:
viewing the net queue.
It ALWAYS shows the wrong cpu numbers.
IE got computers 1,2,3 computer 1 is rendering in slave mode
net queue will show cpu 1 has having failed. I can look over and it
is running ok though. It also will complete it's assignment.
The log shows nothing out of sorts.
As my farm gets bigger geographicly, I hate playing guessing games
is it really down or is is memorex?
On the server when a new slave comes up I get
NET000x.eqp with x matching the slave number
When assigned a task NET000x.cmd,act,bsy are created
Again all numbers match correctly
It seems the number of the slave i++ or i– only when
viewed by net queue on the master.
Anybody else seen this…or have a solution?
Thanks
>> net queue will show cpu 1 has having failed. I can look over and it
is running ok though. <<
Sounds like you need to sync the clocks between systems. For instance, in
Lantastic I use "NET CLOCK \\server-name" to match all system clocks to one
server. Most networs have a similar function.
– Dave
The answer to your problem of having machines be marked as failed even though
they haven't is caused by the slave spending a long time doing something and
not being able to occasionally squawk (say it's still okay and running). There
are settings in your 3DSNET.SET file called NET-SQUAWK, NET-POLL-TIME,
NET-TIMEOUT-CHECK and NET-TIMEOUT. I just personally set all my slaves'
NET-TIMEOUT-CHECK's to "NO". I've a lot of slower machines around here w/
not-great memory amounts, but I'm still able to get a goodly number of frames
off them over a weekend — but they would ALWAYS fail timeouts, and then what
happens is you get 2 (or more) machines rendering the same frames.
Dave