Hi, Am 11.06.2014 um 20:44 schrieb VG:
> Thanks for the information. So when I qlogin I am assigned a compute node > with 1 slot/core(I didn't use the -pe threaded option for more cores/slots). > > The cluster assigns me a node with 8 cores of which I have been assigned 1 > core/slot. Correct. And it would be best not to challenge the admin but stay in the granted limits. > I did qstat -F and got some info for that node which is as follows > > [email protected] IP 0/7/8 0.04 lx26-amd64 > hl:arch=lx26-amd64 > hl:num_proc=16 > hl:mem_total=60.980G > hl:swap_total=996.207M > hl:virtual_total=61.953G > hl:load_avg=0.040000 > hl:load_short=0.040000 > hl:load_medium=0.040000 > hl:load_long=0.000000 > hl:mem_free=59.794G > hl:swap_free=4.000K > hl:virtual_free=59.794G > hl:mem_used=1.186G > hl:swap_used=996.203M > hl:virtual_used=2.159G > hl:cpu=0.300000 > hl:m_topology=SCCCCSCCCCSCCCCSCCCC > hl:m_topology_inuse=SCCCCSCCCCSCCCCSCCCC > hl:m_socket=4 > hl:m_core=16 > hl:np_load_avg=0.002500 > hl:np_load_short=0.002500 > hl:np_load_medium=0.002500 > hl:np_load_long=0.000000 > qf:qname=login > qf:hostname=compute-0-1005.local > qc:slots=1 > qf:tmpdir=/tmp > qf:seq_no=0 > qf:rerun=0.000000 > qf:calendar=NONE > qf:s_rt=infinity > qf:h_rt=infinity > qf:s_cpu=infinity > qf:h_cpu=infinity > qf:s_fsize=infinity > qf:h_fsize=infinity > qf:s_data=infinity > qf:h_data=infinity > qf:s_stack=infinity > qf:h_stack=infinity > qf:s_core=infinity > qf:h_core=infinity > qf:s_rss=infinity > qf:h_rss=infinity > qf:s_vmem=infinity > qf:h_vmem=infinity > qf:min_cpu_interval=00:05:00 > 6141546 0.50500 QLOGIN vgupta12 r 06/11/2014 12:00:27 1 > > Only qlogin will place me in one of the slots/cores out of total 8 cores > here. What I want to know is how much ram is available for me and how much > and how many processor. But i think the node is shared always between cores > so I have to specify some thing like this qloogin -pe threaded 2 -l mf=4G > num_proc=4 and then the cluster will put me on a particular node with the > resources I demanded. Is it correct?? Yep. There are several ways to set up a request for memory. Ask the admin which way he chose. It can be mem_free, virtual_free or h_vmem. Each has certain advantages and side effects. -- Reuti > > Regards > Varun > > > On Wed, Jun 11, 2014 at 2:16 PM, Reuti <[email protected]> wrote: > Hi, > > Am 11.06.2014 um 17:34 schrieb VG: > > > I am working on SGE. > > When I ssh to connect to cluster it connects me to the head node. Then I > > use qlogin to get onto one of the compute nodes/ or lets say one of the > > host machine. > > > > To my understanding I think when I qlogin, the compute node which is > > provided to me , I can use all the resources that machine/compute node has. > > Is there a way I can find out how many cpu processors are available for > > that node and what is the RAM. > > > > Hope to hear from you soon. > > No. You are entitled to use the resources you requested only. In case you > want to use e.g. more than one core, it should be requested when you issue > `qlogin` by requesting a PE. Depending on the Version of SGE, the limit might > even be set on a kernel level, and all of your processes will run on one core > only, leaving the other ones free for other jobs otherwise. > > Memory can for example be set by "-l h_vmem" and checked by `ulimit -Hv` (the > request of "h_vmem" can be set up to be mandatory by the admin). > > -- Reuti > > _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
