E@H on Scientific Linux 6.2?

Forums › Problems and Bug Reports

joe areeda
joe areeda
Joined: 13 Dec 10
Posts: 285
Credit: 320378898
RAC: 0
4 Jun 2012 12:51:08 UTC
Topic 196378
(moderation:
    )

    I'm experimenting with the idea of changing my workstation from Ubuntu to Scientific Linux 6.2 and I'm having trouble running the boinc manager, the client seems to run so if someone could direct me to command line instructions for attaching it to Einstein at Home, my immediate problem goes away.

    I have a clean install of SL6.2, actually it took about 6 of those to figure out how to get NVIDIA drivers to work.

    Then there is a lot of stuff for the LIGO Data Grid reference platform, which I mention because downgrading packages presents a big problem.

    Then to get BOINC manager to run I've added:

    • * GTK to find libwx_gtk (something) sorry I didn't write down the exact name of the missing library
      * OpenSSL 0.9.8 compatibility libraries, fortunately they can coexist with the installed version " OpenSSL 1.0.0-fips 29 Mar 2010"
      * Now I'm stuck on libnotify.so.4, I have:
      Installed Packages
      Name : libnotify
      Arch : x86_64
      Version : 0.5.0
      Release : 1.el6

    I'd appreciate any advice on how to proceed.

    I know this is a BOINC issue not an E@H one but I figured there might be more SL6 with LSCSoft package experience here so I cross posted.

    Joe

    tullio
    tullio
    Joined: 22 Jan 05
    Posts: 2118
    Credit: 61407735
    RAC: 0

    E@H on Scientific Linux 6.2?

    4 Jun 2012 14:55:55 UTC
    Message 110008
    (moderation:
      )

      I am running a BOINC client on a Solaris Virtual Machine without a BOINC manager. If I remember well I used a command line like this:
      boinc_client --attach http://einstein.phys.uwm.edu (in your case) account-key
      Tullio

      joe areeda
      joe areeda
      Joined: 13 Dec 10
      Posts: 285
      Credit: 320378898
      RAC: 0

      Thank you Tullio! For the

      4 Jun 2012 17:11:15 UTC
      Message 110009 in response to message 110008
      (moderation:
        )

        Thank you Tullio!

        For the next guy the program is actually run_client and the option is actually --attach_project easily both easily obtained from ls and --help.

        It seems to be working and detects mouse, keyboard activity.

        Joe

        Bikeman (Heinz-Bernd Eggenstein)
        Bikeman (Heinz-...
        Moderator
        Joined: 28 Aug 06
        Posts: 3522
        Credit: 730679292
        RAC: 1192422

        ....or "boinccmd" on all

        4 Jun 2012 19:06:03 UTC
        Message 110010 in response to message 110009
        (moderation:
          )

          ....or "boinccmd" on all platforms these days.

          I have the same experience with SL and BOINC manager :-(, and didn't care too much to resolve it.

          Cheers HB

          joe areeda
          joe areeda
          Joined: 13 Dec 10
          Posts: 285
          Credit: 320378898
          RAC: 0

          Well, I'm getting closer, I

          4 Jun 2012 22:36:17 UTC
          Message 110011
          (moderation:
            )

            Well, I'm getting closer, I have CPU jobs running and can use the boinc manager from an Ubuntu laptop to see what's going on. I suppose I should try an account manager soon but I really do need to do some of my work.

            I'm not getting any GPU tasks. The log says:

            Mon 04 Jun 2012 03:24:37 PM PDT | Einstein@Home | Sending scheduler request: Requested by user.
Mon 04 Jun 2012 03:24:37 PM PDT | Einstein@Home | Requesting new tasks for NVIDIA
Mon 04 Jun 2012 03:24:39 PM PDT | Einstein@Home | Scheduler request completed: got 0 new tasks
Mon 04 Jun 2012 03:24:39 PM PDT | Einstein@Home | No work sent
Mon 04 Jun 2012 03:24:39 PM PDT | Einstein@Home | see scheduler log messages on http://einstein.phys.uwm.edu//host_sched_logs/5425/5425155
Mon 04 Jun 2012 03:24:39 PM PDT | Einstein@Home | Gamma-ray pulsar search #1 is not available for your type of computer.
Mon 04 Jun 2012 03:24:39 PM PDT | Einstein@Home | Binary Radio Pulsar Search (Arecibo) is not available for your type of computer.


            Which looks an awful lot like "no gpu detected" BUT the web site says "Coprocessors NVIDIA GeForce GTX 560 (1023MB)" but does not list a driver version, which I suspect is the problem. The host_sched_logs don't shed any light on the issue for me.

            nvidia-smi says:

            joe@george:~$ nvidia-smi
            Mon Jun 4 15:14:56 2012
            +------------------------------------------------------+
            | NVIDIA-SMI 3.295.53 Driver Version: 295.53 |
            |-------------------------------+----------------------+----------------------+
            | Nb. Name | Bus Id Disp. | Volatile ECC SB / DB |
            | Fan Temp Power Usage /Cap | Memory Usage | GPU Util. Compute M. |
            |===============================+======================+======================|
            | 0. GeForce GTX 560 | 0000:01:00.0 N/A | N/A N/A |
            | 40% 56 C N/A N/A / N/A | 15% 152MB / 1023MB | N/A Default |
            |-------------------------------+----------------------+----------------------|
            | Compute processes: GPU Memory |
            | GPU PID Process name Usage |
            |=============================================================================|
            | 0. Not Supported |
            +-----------------------------------------------------------------------------+

            That comes from kmod-nvidia which I believe is the current one from the nVidia site also.

            BTW this is CPUID: 5425155

            Any suggestions on where I can look for this issue?

            Joe

            ps. You gentlemen have solved the problem I posted, the forums at boinc have not responded yet.[/code]

            tullio
            tullio
            Joined: 22 Jan 05
            Posts: 2118
            Credit: 61407735
            RAC: 0

            I am running SL in my

            5 Jun 2012 3:44:11 UTC
            Message 110012
            (moderation:
              )

              I am running SL in my BOINC_VM window at Test4Theory@home. There is no GUI and it uses only 256 MB RAM running CERN jobs.
              Tullio

              Claggy
              Claggy
              Joined: 29 Dec 06
              Posts: 560
              Credit: 2699403
              RAC: 0

              Have you got 32bit libaries

              5 Jun 2012 12:12:56 UTC
              Message 110013 in response to message 110011
              (moderation:
                )

                Have you got 32bit libaries installed? as on your last request you're asking for X64 work, not x86 work, and the only work you're done is with the x64 Gravitational Wave S6 LineVeto app:

                All tasks for computer 5425155

                Quote:
                2012-06-05 11:21:44.4469 [PID=18170] Request: [USER#xxxxx] [HOST#5425155] [IP xxx.xxx.xxx.10] client 7.0.25
                Snip
                2012-06-05 11:21:44.5253 [PID=18170] [send] effective_ncpus 8 max_jobs_on_host_cpu 999999 max_jobs_on_host 999999
                2012-06-05 11:21:44.5253 [PID=18170] [send] effective_ngpus 1 max_jobs_on_host_gpu 999999
                2012-06-05 11:21:44.5253 [PID=18170] [send] Not using matchmaker scheduling;2012-06-05 11:21:44.6089 [PID=18188] Request: [USER#7986] [HOST#5021625] [IP 216.41.44.18] client 6.10.58
                Not using EDF sim
                2012-06-05 11:21:44.5253 [PID=18170] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
                2012-06-05 11:21:44.5253 [PID=18170] [send] CUDA: req 21780.00 sec, 1.00 instances; est delay 0.00
                2012-06-05 11:21:44.5253 [PID=18170] [send] work_req_seconds: 0.00 secs
                2012-06-05 11:21:44.5253 [PID=18170] [send] available disk 99.88 GB, work_buf_min 0
                2012-06-05 11:21:44.5253 [PID=18170] [send] active_frac 0.967616 on_frac 0.945005 DCF 1.385627
                2012-06-05 11:21:44.5285 [PID=18170] [send] [HOST#5425155] is reliable
                2012-06-05 11:21:44.5286 [PID=18170] [send] set_trust: error rate 0.100000 > 0.050000, don't trust
                2012-06-05 11:21:44.5407 [PID=18170] [version] no app version available: APP#19 (einsteinbinary_BRP4) PLATFORM#7 (x86_64-pc-linux-gnu) min_version 0
                2012-06-05 11:21:44.5408 [PID=18170] [version] no app version available: APP#17 (hsgamma_FGRP1) PLATFORM#7 (x86_64-pc-linux-gnu) min_version 0
                2012-06-05 11:21:44.5443 [PID=18170] [send] [HOST#5425155] is looking for work from a non-preferred application
                2012-06-05 11:21:44.6475 [PID=18170] [version] Don't need CPU jobs, skipping version 112 for einstein_S6LV1 ()
                2012-06-05 11:21:44.6475 [PID=18170] [version] Checking plan class 'X64'
                2012-06-05 11:21:44.6479 [PID=18170] [version] reading plan classes from file '../plan_class_spec.xml'
                2012-06-05 11:21:44.6479 [PID=18170] [version] Don't need CPU jobs, skipping version 113 for einstein_S6LV1 (X64)
                2012-06-05 11:21:44.6480 [PID=18170] [version] no app version available: APP#20 (einstein_S6LV1) PLATFORM#7 (x86_64-pc-linux-gnu) min_version 0
                2012-06-05 11:21:44.6496 [PID=18170] [debug] [HOST#5425155] MSG(high) No work sent
                2012-06-05 11:21:44.6496 [PID=18170] [debug] [HOST#5425155] MSG(high) Binary Radio Pulsar Search (Arecibo) is not available for your type of computer.
                2012-06-05 11:21:44.6496 [PID=18170] [debug] [HOST#5425155] MSG(high) Gamma-ray pulsar search #1 is not available for your type of computer.
                2012-06-05 11:21:44.6496 [PID=18170] [debug] [HOST#5425155] MSG(high) see scheduler log messages on http://einstein.phys.uwm.edu//host_sched_logs/5425/5425155
                2012-06-05 11:21:44.6496 [PID=18170] Sending reply to [HOST#5425155]: 0 results, delay req 60.00
                2012-06-05 11:21:44.6499 [PID=18170] Scheduler ran 0.210 seconds

                Claggy

                joe areeda
                joe areeda
                Joined: 13 Dec 10
                Posts: 285
                Credit: 320378898
                RAC: 0

                Claggy, I was just

                5 Jun 2012 21:56:51 UTC
                Message 110014 in response to message 110013
                (moderation:
                  )

                  Claggy,

                  I was just thinking about that on my way home from work today. Good catch.

                  Any idea what the 32bit librarys are called on SL6 (fedora, centos, redhat relative)? ia32 is not it.

                  Joe

                  joe areeda
                  joe areeda
                  Joined: 13 Dec 10
                  Posts: 285
                  Credit: 320378898
                  RAC: 0

                  Well installing

                  6 Jun 2012 1:59:26 UTC
                  Message 110015 in response to message 110014
                  (moderation:
                    )

                    Well installing nvidia-x11-drv-32bit seems to allow CUDA tasks to run but they all error out with:

                    7.0.25

                    process exited with code 127 (0x7f, -129)

                    ../../projects/einstein.phys.uwm.edu/einsteinbinary_BRP4_1.24_i686-pc-linux-gnu__BRP4cuda32nv270: error while loading shared libraries: libstdc++.so.6: wrong ELF class: ELFCLASS64

                    ]]>

                    I'm not sure there's anything to do about that. I don't think I can install 32 bit versions of c++ libraries.

                    This machine was getting 30K credits/day a bit part of my RAC and I'm building a couple of systems that should beat it. At least one will be running Windows, but the other needs to run SL6 which is the main reason this one was changed. I was also sick and tired of Ubuntu's bugginess and hope SL6 is more stable.

                    Joe

                    joe areeda
                    joe areeda
                    Joined: 13 Dec 10
                    Posts: 285
                    Credit: 320378898
                    RAC: 0

                    OK this is kind of

                    19 Jul 2012 16:56:43 UTC
                    Message 110016 in response to message 110015
                    (moderation:
                      )

                      OK this is kind of weird.

                      I checked today for the first time in a while and this system is now processing and validating BRP4 cuda tasks using the GTX560 in the system.

                      I'm not sure if we have some new software coming in or one of the updates fixed something.

                      Has anyone else seen a sudden ability to use GPUs on SL6.2?

                      On July 7 every one was producing errors eg: http://einsteinathome.org/workunit/127017261 and http://einsteinathome.org/workunit/127005019

                      Then on July 8 they all seem to validate eg: http://einsteinathome.org/workunit/127061065 and http://einsteinathome.org/workunit/127061073

                      The real embarrassing part is that I thought I had set this system not to use GPU, but my expertise with boinccmd is severely lacking

                      Joe

                      joe areeda
                      joe areeda
                      Joined: 13 Dec 10
                      Posts: 285
                      Credit: 320378898
                      RAC: 0

                      I have a bit more to

                      23 Jul 2012 1:43:14 UTC
                      Message 110017
                      (moderation:
                        )

                        I have a bit more to report.

                        I spent the better part of a Sunday compiling the boinc client from sources, with help from the boinc mailing list.

                        I now have 7.0.31 running under SL6 client and manager. It's not quite perfect yet but I've got 3 BRP4 tasks waiting validation and the einstein preferences allow 2 tasks on the GPU.

                        The open issues are:

                        The manager is looking in my home directory for boinc whereas the make install put it in /usr/local/bin. So I have to start the client then the manager.

                        I didn't save the init.d stuff that makes the client a service loaded on boot.

                        But I'm pretty content with it the way it is.

                        Joe

                        Comment viewing options

                        Select your preferred way to display the comments and click "Save settings" to activate your changes.

                        Forums › Problems and Bug Reports