Hi Flemming, Here is an example on rcas0032. the total iowait > 20 % Usually both cpu's are fully loaded and one cannot do anything. Selemon, ===================================================================== tigist_at_rcas0032> top 11:02:22 up 67 days, 1:56, 1 user, load average: 2.86, 2.66, 2.60 87 processes: 86 sleeping, 1 running, 0 zombie, 0 stopped CPU states: cpu user nice system irq softirq iowait idle total 0.9% 5.9% 1.9% 0.0% 0.4% 52.7% 37.9% cpu00 1.9% 2.9% 1.9% 0.0% 0.9% 22.7% 69.3% cpu01 0.0% 8.8% 1.9% 0.0% 0.0% 82.3% 6.8% Mem: 1025428k av, 1008760k used, 16668k free, 0k shrd, 7452k buff 784808k actv, 148544k in_d, 13724k in_c Swap: 1574328k av, 787420k used, 786908k free 54388k cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 19893 chiu 26 10 697M 596M 1704 D N 5.9 59.5 71:22 1 root.exe 7327 root 23 0 1552 1552 1028 S 0.9 0.1 0:00 1 id -----Original Message----- From: Flemming Videbaek [mailto:videbaek_at_bnl.gov] Sent: Thu 2/1/2007 5:43 PM To: Bekele, Selemon Cc: devlist Subject: Re: [Brahms-dev-l] RCAS status Well, This sounds it is not a local problem. I.e. it may be related to netwrok access, and your jobs rying to access certain files. Is it that the jobs gets no cpu time? The foreign condor jobs are certainly running a a high nice level and ought not to cause problems. If you want we can schedule a test when you do interactive work. I can kill foreign condor-jobs on specficic machines, though it is not so nice to do, and see for a test if this is the cause or what? Flemming -------------------------------------------- Flemming Videbaek Physics Department Bldg 510-D Brookhaven National Laboratory Upton, NY11973 phone: 631-344-4106 cell: 631-681-1596 fax: 631-344-1334 e-mail: videbaek @ bnl gov ----- Original Message ----- From: "Bekele, Selemon" <bekeleku_at_ku.edu> To: "Flemming Videbaek" <videbaek_at_bnl.gov> Sent: Thursday, February 01, 2007 3:47 PM Subject: RE: [Brahms-dev-l] RCAS status Hi, this has actually been a problem for the last couple of days. It was virtually impossible to run any interactive jobs on may of brahms nodes. I have been using rcas0032, rcas0034. selemon, -----Original Message----- From: brahms-dev-l-bounces_at_lists.bnl.gov on behalf of Flemming Videbaek Sent: Thu 2/1/2007 12:53 PM To: hongyan_at_ift.uib.no; devlist Subject: Re: [Brahms-dev-l] RCAS status Hi This has been the policy of having general queues running on our clusters for quite a while. Unless the jobs are special this in general has a very small impact, since they are niced; our condor jobs have priority. Thus give me specific information on what machines seems to be slow, so one can look at the performance. One easy way is to do a top - check for io/wait. If this is large (>20% there is a potential problem) - CTRL M ->order jobs by memory usage. There may be very large jobs doint little. if any of these conditions are fullfield there may be a problem. If not it is perception, not a real impact (95% confidence level) /fv -------------------------------------------- Flemming Videbaek Physics Department Bldg 510-D Brookhaven National Laboratory Upton, NY11973 phone: 631-344-4106 cell: 631-681-1596 fax: 631-344-1334 e-mail: videbaek @ bnl gov ----- Original Message ----- From: "Hongyan Yang" <hongyan_at_ift.uib.no> To: "devlist" <brahms-dev-l_at_lists.bnl.gov> Sent: Thursday, February 01, 2007 1:31 PM Subject: [Brahms-dev-l] RCAS status > Dear all, > > I am wondering if any of you've noticed that almost all of our RCAS > machines are busy with some condor jobs (not from BRAHMS) - which > affected the speed of a terminal to complete any local session. This has > been for quite a while - is there any restriction for people from other > collaboration to use our cluster? > > Hope this can be solved (or at least compromised), to make sure at least > we have enough machines to do our jobs. > > Best regards, > Hongyan > > -- > Hongyan YANG > Department of Physics and Technology Phone: +47 55 58 27 25 (Office) > University of Bergen > Allegt. 55, 5007 Bergen Fax : +47 55 58 94 40 > Norway Email: hongyan[at]ift.uib.no > > > _______________________________________________ > Brahms-dev-l mailing list > Brahms-dev-l_at_lists.bnl.gov > http://lists.bnl.gov/mailman/listinfo/brahms-dev-l > _______________________________________________ Brahms-dev-l mailing list Brahms-dev-l_at_lists.bnl.gov http://lists.bnl.gov/mailman/listinfo/brahms-dev-l _______________________________________________ Brahms-dev-l mailing list Brahms-dev-l_at_lists.bnl.gov http://lists.bnl.gov/mailman/listinfo/brahms-dev-lReceived on Fri Feb 02 2007 - 11:09:33 EST
This archive was generated by hypermail 2.2.0 : Fri Feb 02 2007 - 11:09:58 EST