Advanced search

Message boards : Number crunching : no checkpoints

AuthorMessage
Michael Karlinsky
Avatar
Send message
Joined: Aug 3 10
Posts: 10
Credit: 20,141
RAC: 0
Message 72 - Posted 4 Aug 2010 8:00:25 UTC

    Last modified: 4 Aug 2010 8:01:07 UTC

    Hi,

    after booting up my machine in the morning :) I noticed, that computation starts from scratch.

    So I guess the app does not checkpoint? OK, tasks are short, but nontheless. Any plans to implement that. (If time permits, that is.)


    Michael
    ____________
    Team Linux Users Everywhere

    Fabio Campos
    Volunteer moderator
    Project administrator
    Project developer
    Project tester
    Volunteer developer
    Volunteer tester
    Project scientist
    Send message
    Joined: Feb 3 10
    Posts: 65
    Credit: 10,767
    RAC: 11
    Message 73 - Posted 4 Aug 2010 12:00:00 UTC - in response to Message 72.

      Last modified: 4 Aug 2010 12:00:12 UTC

      Hi Michael,

      that's not right. the primaboinca-app does (actually too much; after every crunched number) checkpoint. so I can't understand, why from scratch ...

      John McLeod VII
      Avatar
      Send message
      Joined: Aug 4 10
      Posts: 24
      Credit: 521,799
      RAC: 377
      Message 75 - Posted 4 Aug 2010 12:37:35 UTC - in response to Message 73.

        Hi Michael,

        that's not right. the primaboinca-app does (actually too much; after every crunched number) checkpoint. so I can't understand, why from scratch ...

        If possible, you should pay attention to the flag the BOINC daemon passes you indicating when you should checkpoint. If it takes a little to get to the next checkpoint, that is OK.

        It is up to the application to read the checkpoint file that it wrote and continue on from there.
        ____________


        BOINC WIKI

        Truth?
        Send message
        Joined: Jul 25 10
        Posts: 1
        Credit: 1,206,476
        RAC: 1,122
        Message 89 - Posted 5 Aug 2010 20:54:57 UTC

          I also have been having trouble with some tasks they pause after the 1 hour of work( which works fine with other projects). With this project if it takes longer than 1 hour it pauses (normal), it will start back a 0% complete (abnormal). It will cycle through this until, BOINC determines it cannot be finished in the allotted time and goes into panic mode. When it goes to panic mode I am able to compete the task.

          Does this just happen with certain tasks? Is there something I can do to avoid this? Any information would be greatly appreciated.

          Thank you for your time,

          John McLeod VII
          Avatar
          Send message
          Joined: Aug 4 10
          Posts: 24
          Credit: 521,799
          RAC: 377
          Message 91 - Posted 5 Aug 2010 23:04:50 UTC - in response to Message 89.

            I also have been having trouble with some tasks they pause after the 1 hour of work( which works fine with other projects). With this project if it takes longer than 1 hour it pauses (normal), it will start back a 0% complete (abnormal). It will cycle through this until, BOINC determines it cannot be finished in the allotted time and goes into panic mode. When it goes to panic mode I am able to compete the task.

            Does this just happen with certain tasks? Is there something I can do to avoid this? Any information would be greatly appreciated.

            Thank you for your time,

            This is probably because the project is telling BOINC that it has checkpointed when in reality the application has not stored away its state in a form the application can read it back in.

            BOINC will save a task to RAM if it has not checkpointed when the task is swapped out. BOINC will also save a task to RAM if the user has set the appropriate setting. Otherwise, BOINC removes the task from RAM.
            ____________


            BOINC WIKI

            Fabio Campos
            Volunteer moderator
            Project administrator
            Project developer
            Project tester
            Volunteer developer
            Volunteer tester
            Project scientist
            Send message
            Joined: Feb 3 10
            Posts: 65
            Credit: 10,767
            RAC: 11
            Message 99 - Posted 6 Aug 2010 7:15:54 UTC - in response to Message 91.

              Hi,

              I'll check this ...

              John McLeod VII
              Avatar
              Send message
              Joined: Aug 4 10
              Posts: 24
              Credit: 521,799
              RAC: 377
              Message 102 - Posted 6 Aug 2010 12:38:04 UTC - in response to Message 99.

                Hi,

                I'll check this ...

                When an application starts, it has to test to see if there is a checkpoint file. If there is one, the application has to load it on start.
                ____________


                BOINC WIKI

                Profile JHAPA
                Avatar
                Send message
                Joined: Aug 11 10
                Posts: 3
                Credit: 2,670,105
                RAC: 2,288
                Message 145 - Posted 23 Aug 2010 11:00:16 UTC - in response to Message 102.

                  Hi all, i have the same problem, i looked into slots and no checkpoint file find here...so that means, if WU is stoped , then no information about progres is saved...WU has maybe checkpoints, but WU is not able to write it...my opinion was that is problem of Boinc client (a had 6.10.56 and not bad but not excelent) but after update to 6.11.6 is the same...my second thinking is about combination of access rights to write in folder for WU and W7 64bit disk management, that means, that WU hasent enough rights to write this file into Aplication Data folder...Or it is problem of WU, that is bug in saving of checkpoint file?
                  ____________
                  Member of PrimeSearchTeam

                  skildude
                  Send message
                  Joined: Sep 1 10
                  Posts: 1
                  Credit: 106,425
                  RAC: 20
                  Message 162 - Posted 2 Sep 2010 15:30:47 UTC - in response to Message 145.

                    Noted the problem and also noted that the clock time didnt restart with the WU. I nearly doubled my credit depending on how deep I was into the WU.

                    Supraprime
                    Send message
                    Joined: Sep 6 10
                    Posts: 5
                    Credit: 1,080,300
                    RAC: 895
                    Message 167 - Posted 8 Sep 2010 7:53:33 UTC

                      Hi all.
                      I also have the same prob on my Vista 64,Bouinc 6.10.56.After restart all wu's start from 0%.
                      Greetings from Germany.

                      Fabio Campos
                      Volunteer moderator
                      Project administrator
                      Project developer
                      Project tester
                      Volunteer developer
                      Volunteer tester
                      Project scientist
                      Send message
                      Joined: Feb 3 10
                      Posts: 65
                      Credit: 10,767
                      RAC: 11
                      Message 168 - Posted 8 Sep 2010 10:52:05 UTC - in response to Message 167.

                        Hi everybody,

                        I'm working on it ...

                        Matthias Lehmkuhl
                        Send message
                        Joined: Sep 22 10
                        Posts: 2
                        Credit: 1,029,005
                        RAC: 439
                        Message 181 - Posted 29 Sep 2010 19:51:42 UTC - in response to Message 168.

                          Hi everybody,

                          I'm working on it ...

                          Hi,
                          any new status?
                          Have the same behavior (Boinc 5.10.58), time will count from last state,
                          but result runs from start.
                          ____________
                          Matthias

                          sonthakit
                          Send message
                          Joined: Sep 1 10
                          Posts: 1
                          Credit: 63,466
                          RAC: 22
                          Message 192 - Posted 9 Oct 2010 12:39:15 UTC - in response to Message 168.

                            Hi everybody,

                            I'm working on it ...


                            One month had pass. Nothing change.
                            This my WU is just only 17 min left, it reset to 0% again when window restart...
                            :(

                            whynot
                            Send message
                            Joined: Sep 15 10
                            Posts: 30
                            Credit: 10,790,844
                            RAC: 2,059
                            Message 193 - Posted 9 Oct 2010 13:43:45 UTC - in response to Message 91.


                              BOINC will save a task to RAM if it has not checkpointed when the task is swapped out. BOINC will also save a task to RAM if the user has set the appropriate setting. Otherwise, BOINC removes the task from RAM.


                              Applied as a temporal workaround. That 'temporal' is a bit optimistic, isn't it? Unluckily, that setting is host-wide, at least, and applies to all projects running. But it improves things a lot anyway. Thus I don't care.

                              ChertseyAl
                              Avatar
                              Send message
                              Joined: Jul 28 10
                              Posts: 57
                              Credit: 6,881,902
                              RAC: 2,269
                              Message 194 - Posted 10 Oct 2010 16:47:48 UTC - in response to Message 73.

                                the primaboinca-app does (actually too much; after every crunched number) checkpoint.


                                Seems to be writing to disk about once per second. A couple of my machines have naturally 'noisy' disks, so it's like being in a room full of crickets :(

                                I'll stop crunching this project on those (if not all) machines.

                                Al.

                                whynot
                                Send message
                                Joined: Sep 15 10
                                Posts: 30
                                Credit: 10,790,844
                                RAC: 2,059
                                Message 198 - Posted 16 Oct 2010 12:32:06 UTC

                                  If my reading of this is correct then other approach would be setting 'Switch between applications' to 120min (or 90min at least). However, me thinks, that checkpoint issue should be fixed app-side anyway. Pity.

                                  ____________
                                  I'm counting for science.
                                  Points just make me sick.

                                  Charles Elliott
                                  Send message
                                  Joined: Oct 28 10
                                  Posts: 2
                                  Credit: 61,233
                                  RAC: 0
                                  Message 205 - Posted 1 Nov 2010 1:52:14 UTC - in response to Message 73.

                                    PrimaBoinca is not checkpointing. I have 32 active tasks. I have watched Boinc restart 16 of them. Every one went from 80-99% done to 0% done when restarted.
                                    ____________

                                    Post to thread

                                    Message boards : Number crunching : no checkpoints


                                    Main page · Your account · Message boards


                                    Copyright © 2017 primaboinca.com