Hi.
To the advantage of others, I post here my solution to this problem.
It may not be the optimal, but now the jobs are flowing and getting running.
It turns out that if 'mem_free' is configured as a consumbale resource,
you need to specify the 'mem_free' capacity for your execution hosts in
'complex_values'.. so, I just added mem_free=XX.XG for each execution host.
Even though I'm pretty sure that my configuration has always been with
mem_free as a consumable resource and mem_free never declared for each
host complexes, something may have altered the configuration when I was
playing with various other configuration parameters.
Best regards.
Robi
Il 13.05.2014 19:05, Roberto Nunnari ha scritto:
Hi.
I must have broken my scheduler configuration.. it doesn't recognize
anymore the -l mem_free flag.. well.. I guess it checks the flag at
submit time, but when it tries to schedule it, it shows the following
output:
$ qstat -j 1637743 |grep "(mem_free)"
(-l h_rt=864000,mem_free=3072M) cannot run
in queue "[email protected]" because job requests unknown resource
(mem_free)
...
$ qconf -sc | grep mem_free
mem_free mf MEMORY <= YES YES 0 0
$ qconf -se node09 | grep mem_free
mem_free=15558.878906M,swap_free=0.000000M, \
Any ideas? How to fix this? What should I check? What have I broken?
Thank you and best regards.
Robi
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
