Message boards : Number crunching : New work Discussion
Message board moderation
Previous · 1 . . . 63 · 64 · 65 · 66 · 67 · 68 · 69 . . . 91 · Next
Author | Message |
---|---|
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
OK. My last batch 901 has finished: First was the restart file, which took 3 seconds to upload Then a 20 second gap Then the trickle_up file was sent Another 30 second gap The zip 4 was uploaded at about 32 minutes before completion. This took about 3.5 minutes to upload. I missed all of these, so I don't know about sizes. About 20 minutes later the task finished, and the out file was uploaded - 238.04 It happened so fast that I didn't see whether that was bytes or kilobytes. :( So, if there's "no congestion at the traffic lights", there's plenty of time for each file to get through before the next part shows up. :) |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,024,725 RAC: 20,592 |
Now instead of exceeding a file size you're talking about how many files are being uploaded at the same time. I'm now running 3,201 WUs of various projects so that will be next to impossible. The thing that needs changing is <max_nbytes>150000000.000000</max_nbytes> in client_state.xml but unless you have a slow or unreliable connection or i realise what caused the problem in my case, turning off internet access for BOINC to allow other half to have a Zoom meeting then this shouldn't be a problem. I don't recommend playing with this file if you don't need to and if you do, please do make a back up. I don't turn BOINC off before editing it but I do stop computation on all running tasks. The issue has cropped up a time or two in testing but any relevant testing for these obviously didn't include any of us turning off internet access at a critical time. Edit: Another option of course is to just pause any task that is close to finishing when internet access might be uncompromised. I think from our discussions Sarah is going to increase the limit for the third batch of tasks from this experiment. so on that one it shouldn't be an issue. |
Send message Joined: 1 Jan 07 Posts: 1061 Credit: 36,708,278 RAC: 9,361 |
The thing that needs changing isThat makes sense, but I disagree with the bit I've marked in red. The BOINC client writes out a whole new copy of client_state.xml every minute or so (and keeps it's own backup as client_state_prev.xml), but it only ever reads it as the client is starting. While the client is running, the whole darn thing is kept in memory, and if you try to edit the file copy, your changes will be over-written and lost at the next write. You need to stop the client (only - can keep the manager open) before starting to change anything. Oh, and use a plain-text editor - BOINC doesn't like the output written by 'clever' XML tools. |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,024,725 RAC: 20,592 |
That makes sense, but I disagree with the bit I've marked in red. Oops. As always you are right Richard. Shows that it has been a while since I needed to do this. |
Send message Joined: 5 Aug 04 Posts: 1120 Credit: 17,202,915 RAC: 2,154 |
Out of idle curiosity, I looked at client-state.xml max_nbytes and got; $ grep max_nbytes /var/lib/boinc/client_state.xml | wc -l 430 430 entries! Thank goodness I do not seem to need to change one of them. How would I figure out which one needs changing? Four of them looked like this: <max_nbytes>0.000000</max_nbytes> <max_nbytes>0.000000</max_nbytes> <max_nbytes>104857600.000000</max_nbytes> <max_nbytes>104857600.000000</max_nbytes> |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
It's the upload zip for your current task name. |
Send message Joined: 5 Aug 04 Posts: 1120 Credit: 17,202,915 RAC: 2,154 |
So, if there's "no congestion at the traffic lights", there's plenty of time for each file to get through before the next part shows up. :) Seems that way to me too. Wed 27 Oct 2021 04:49:25 AM EDT | climateprediction.net | Started upload of hadam4h_h1bc_200602_4_920_012116922_0_r1467636988_restart.zip Wed 27 Oct 2021 04:49:28 AM EDT | climateprediction.net | Finished upload of hadam4h_h1bc_200602_4_920_012116922_0_r1467636988_restart.zip Wed 27 Oct 2021 04:49:58 AM EDT | climateprediction.net | Sending scheduler request: To send trickle-up message. Wed 27 Oct 2021 04:50:30 AM EDT | climateprediction.net | Started upload of hadam4h_h1bc_200602_4_920_012116922_0_r1467636988_4.zip Wed 27 Oct 2021 04:50:50 AM EDT | climateprediction.net | Finished upload of hadam4h_h1bc_200602_4_920_012116922_0_r1467636988_4.zip Wed 27 Oct 2021 05:11:02 AM EDT | climateprediction.net | Started upload of hadam4h_h1bc_200602_4_920_012116922_0_r1467636988_out.zip Wed 27 Oct 2021 05:11:03 AM EDT | climateprediction.net | Computation for task hadam4h_h1bc_200602_4_920_012116922_0 finished Wed 27 Oct 2021 05:11:06 AM EDT | climateprediction.net | Finished upload of hadam4h_h1bc_200602_4_920_012116922_0_r1467636988_out.zip |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,024,725 RAC: 20,592 |
And having done a search and replace with all the relevant zips and allowed them double the allowance, I can confirm that my task that has just finished is uploading without issue. Always good to get confirmation that I haven't made a mistake editing that file. :) |
Send message Joined: 22 Feb 06 Posts: 491 Credit: 30,994,950 RAC: 14,359 |
It looks as if some of the batches are set to 150000000 any way:- <name>hadam4h_h01m_200702_4_920_012115276_1_r113933754_4.zip</name> <nbytes>0.000000</nbytes> <max_nbytes>150000000.000000</max_nbytes> and batch 895 is set to 210000000! Does anything need to be changed? |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,024,725 RAC: 20,592 |
It looks as if some of the batches are set to 150000000 any way:- My one that failed was set to 150000000. The zip4 is produced slightly before the task finishes and for most will upload before it finishes and reaches the check that makes it fail. This will only cause a problem if upload speeds are very slow, or internet access for BOINC is turned off or down for another reason. The zip4 size reported in the transfer tab is about 147MB binary making it slightly above the 150000000 limit but the file size check never occurs if it is uploaded before the task ends. So for most of us, most of the time, it isn't a problem. |
Send message Joined: 9 Oct 04 Posts: 82 Credit: 69,923,532 RAC: 8,011 |
Just had two with the file size error... at least I understand now, what you were talking about. And yes, I have a slow internet connection. |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
OK, I've just sent an email. I also mentioned the extended error messages . |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
For my batch 921, zip 1 was 150.06Mb That's sailing too close to the wind for safety. I don't know yet what the project is going to do about these. |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,024,725 RAC: 20,592 |
I don't know yet what the project is going to do about these. This is what I got on the task noticeboard page we have access to. Thank you for this. I will update the limit for future batches. This shouldn’t be fundamentally different in size to the other CDDHDD batches But, that was before anyone else posted that they had problems and before you noticed the size of your zip. |
Send message Joined: 5 Aug 04 Posts: 1120 Credit: 17,202,915 RAC: 2,154 |
I do not watch to see the sizes of the uploaded files, but I do know the maximum speed my Internet link can go, that is about 75 MegaBits/second. Sometimes I see it slightly faster. So let us say it is 10 MegaBytes per second. Of coure it can be slower if the Internet is conjested. So here are the latest uploads. Thu 28 Oct 2021 02:03:43 AM EDT | climateprediction.net | Started upload of hadam4h_h0c7_200602_4_920_012115657_0_r77250837_restart.zip Thu 28 Oct 2021 02:03:46 AM EDT | climateprediction.net | Finished upload of hadam4h_h0c7_200602_4_920_012115657_0_r77250837_restart.zip [ 3 seconds so 30 Megabytes estimated.] Thu 28 Oct 2021 02:04:14 AM EDT | climateprediction.net | Sending scheduler request: To send trickle-up message. Thu 28 Oct 2021 02:04:47 AM EDT | climateprediction.net | Started upload of hadam4h_h0c7_200602_4_920_012115657_0_r77250837_4.zip Thu 28 Oct 2021 02:05:46 AM EDT | climateprediction.net | Finished upload of hadam4h_h0c7_200602_4_920_012115657_0_r77250837_4.zip [59 seconds, so 590 Megabytes estimated.] Thu 28 Oct 2021 02:25:48 AM EDT | climateprediction.net | Computation for task hadam4h_h0c7_200602_4_920_012115657_0 finished Thu 28 Oct 2021 02:25:50 AM EDT | climateprediction.net | Started upload of hadam4h_h0c7_200602_4_920_012115657_0_r77250837_out.zip Thu 28 Oct 2021 02:25:53 AM EDT | climateprediction.net | Finished upload of hadam4h_h0c7_200602_4_920_012115657_0_r77250837_out.zip [3 seconds, so 30 Megabytes/second.] Now these speeds are the fastest I could be getting. If, instead, I look at the speeds CPDN says I am getting from this machine, they say Average upload rate 1731.78 KB/sec that is about 6 times slower than I assumed above, so those uploads are more likely to be 5 Megabytes, 100 Megabytes, and 5 Megabytes, respectively. Here are some speed tests of my machine to the Internet. The test server is about 60 miles from me. Timestamp Download Upload Test Server 10/29/2021 10:35:51 76.02 Mbps 89.09 Mbps New York City, NY 10/25/2021 16:52:39 75.81 Mbps 60.33 Mbps New York City, NY Now for this work unit, CPDN reports Peak working set size 1,380.64 MB Peak swap size 1,398.43 MB Peak disk usage 12.91 MB Since I have 65 GBytes of RAM, I have no trouble running several of these. If my peak disk usage is only 13 Megabytes and I even uploaded it all, this does not seem to be very much. |
Send message Joined: 7 Aug 04 Posts: 2187 Credit: 64,822,615 RAC: 5,275 |
The peak disk usage is the largest size of the slots directory that task is using. In most boinc projects, this is where the file reading and writing is being done. However, cpdn uses the ..../projects/climateprediction.net/task name directory for it's reading and writing. |
Send message Joined: 5 Aug 04 Posts: 1120 Credit: 17,202,915 RAC: 2,154 |
OK: here are my disk usages in the relevant directories. directory . is /var/lib/boinc Of course, these are just what is there now. 18232M . 13825M ./projects 11332M ./projects/climateprediction.net 4404M ./slots 3008M ./projects/climateprediction.net/hadam4h_11cx_209902_4_921_012119079 2567M ./projects/climateprediction.net/hadam4h_b0sx_201211_5_882_012036049 2262M ./projects/climateprediction.net/hadam4h_h12y_200902_4_920_012116620 2151M ./projects/climateprediction.net/hadam4h_10x3_209602_4_921_012118509 1905M ./projects/climateprediction.net/hadam4h_b0sx_201211_5_882_012036049/datain 1694M ./projects/climateprediction.net/hadam4h_11cx_209902_4_921_012119079/datain 1694M ./projects/climateprediction.net/hadam4h_10x3_209602_4_921_012118509/datain 1691M ./projects/climateprediction.net/hadam4h_h12y_200902_4_920_012116620/datain 1547M ./projects/boinc.bakerlab.org_rosetta 1518M ./projects/climateprediction.net/hadam4h_b0sx_201211_5_882_012036049/datain/ancil 1313M ./projects/climateprediction.net/hadam4h_11cx_209902_4_921_012119079/dataout 1306M ./projects/climateprediction.net/hadam4h_11cx_209902_4_921_012119079/datain/ancil 1306M ./projects/climateprediction.net/hadam4h_10x3_209602_4_921_012118509/datain/ancil 1304M ./projects/climateprediction.net/hadam4h_h12y_200902_4_920_012116620/datain/ancil |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,024,725 RAC: 20,592 |
they say Mine peaks at about 90KB/sec on bored band. about twice that if I tether my phone and use 4G. Even then, most of the time, the upload happens quickly enough not to get checked and cause a failure. The problem comes if there is a lot of other internet activity at the same time such as other half and possibly myself having zoom meetings at the same time and other concurrent uploads. Sadly, though it has been asked for a number of times, BOINC does not allow selective pausing of uploads/downloads. |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
What is on your HDs isn't important. It's the size of the zip, and once it's uploaded, there's no record of that. So you need to Suspend net access, wait until a zip is created, and then write it down from the Transfers tab. Then Resume net access. I'm in contact with the researcher about this, but haven't heard back from my 3rd email. Oh, and it's the weekend again. :( |
Send message Joined: 5 Aug 04 Posts: 1120 Credit: 17,202,915 RAC: 2,154 |
What is on your HDs isn't important. It's the size of the zip, and once it's uploaded, there's no record of that. I believe am beginning to understand. I am not running out of disk space for these upload files. It is the upload server that is unwilling to accept some (the too-large) of them. But maybe my boinc-client is running a check to prevent my sending too-large .zip files to the upload server. Do I have this right? |
©2024 cpdn.org