[torqueusers] rmcfg madness

Sam Rash srash at yahoo-inc.com
Fri Feb 2 03:29:24 MST 2007


Hey all

 

So suppose I put these lines in my maui.cfg

 

RMCFG[0] AUTHTYPE=CHECKSUM EPORT=15004 TIMEOUT=00:00:09 TYPE=PBS

RMCFG[hostp1] AUTHTYPE=CHECKSUM EPORT=15004 HOST=hostp1 TIMEOUT=00:00:09
TYPE=PBS

 

I know for a fact it messes things up.somehow jobs that get to maui get a
partition req of 'ALL' which as I follow is virtual and is an empty
hostlist, so they don't run.

And it also seems clear to me that putting those 2 lines, 2 rmcfg lines
period is nonsense since maui works with ONE RM.   But can anyone explain
what maui SHOULD do here to handle the user fubar?   Should statement take
precedence?  If so, which and why?   Or should it 'merge' them?  Or is the
behavior totally undefined?

 

I can say for sure

 

1)       putting both in the order you see there causes jobs to never run
(partition=ALL => hostset empty, saw this msg in maui log)

2)       putting just the 2nd line,which is what we want, works

 

but I'm curious if it'd matter (too late here to ttest, and face it I fixed
it)

1)       does the order matter?  Does the lexicographical order of the name
in the RMCFG[.] matter?  Ie maui takes the 'first' or 'last' one and
therefore it's not that there were two RMCFG lines, BUT THE FIRST [0] was
bad and the good 2nd one was ignored

2)       or, it's just that putting two of these in any order is asking for
undefined order

 

 

bottom-line: we never meant to do this, somehow it got into some prototyping
and slipped past several sets of eyes.

 

Thanks in advance for any info ppl

 

-sr

 

 

Sam Rash

srash at yahoo-inc.com

408-349-7312

vertigosr37

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20070202/f144045a/attachment.html


More information about the torqueusers mailing list