LHC@home

I'm sorry Nils, but this is by no means resolved.

I'm still living the problem and the fact that it's reproduced at multiple participants means that it is either related to Boinc or VirtualBox.

For information, BOINC version 7.10.2 / 3.0.1
VirtualBox 5.2.12 r 122591

I'm running Windows 10 Pro version 1709, build 16299.431, patched to the latest updates.
Motherboard Asus A-99 vII
CPU Intel Core i7-6850K @3.6 GHz
Memory Corsair 32 GB
GPU's : Asus Strix 1070Ti 8 GB and 1070 (will have 2 Ti's this weekend)

No overclocking is happening.

I'm running for years LHC on CPU and GPUGrid on 2 GPU's with little issues.
Alternatively I also run ClimatePrediction, Rosetta, WorldCommunityGrid, MilkyWay.

I dedicate 8 out of 12 of the virtual cores to Boinc (reserving 2 of 8 to GPUGrid to manage the GPU WU's),
and rest is shared by LHC and WCG, with priority 2000 to LHC and 150 to WCG to run primarily LHC.
Atlas 5 cores taking .... 5 cores and the last core being shared between LHCb, Theory, CMS and Sixtracks.

As you can see from my stats probably, this has been running quite well, racking up between 5.000 and 10.000 credits
per day.
I only have troubles up-and downloading LHC WU's.

What I did notice, is that when I stop BOINC (after suspending active WU's of course), I exit it and start the
application again, the down & uploads suddenly work again and reach 3.5 Mbit/s speeds, finishing them in
a matter of a few tens of seconds.

Sorry to push but could you please investigate further as this is quite annoying :-S

Many thanks in advance !

K.

[/img]d:\LHC- ATLAS DOWNLOADS.jpg[img][/img]

Nils,

sorry, but the Atlas-download problems are really existend and have already driven crunchers to other projects. Me too !

It looks as if the problem is on the way outside to the world.

If it would help you, you can contact me and we open a remote-session and you can see the problem live.

If I should make a shot I would guess it is a damaged / faulty Switch. I had one in my company and the behaviour was very similar

Yeti

---

Supporting BOINC, a great concept !

maeax wrote:

read this message in this thread:
https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4741&postid=35632#35632
Cern-It is working on this problem.

I don't see that they are working on it. They even can't see the problem

Nils Høimyr wrote:

Download and upload times are like usual on our BOINC clients.

Could this be local, or have been a temporary storage issue? As far as we can see from our monitoring data, I/O is normal over the last 7 days.

+1

---

Sorry, but as mentioned, on our BOINC clients upload speed is normal.

Its definitely a problem. This just started happening to both downloads and uploads. I have over 20 results trying to upload that keep failing. A week ago I never saw this issue. There is something going on whether you personally see it on your end or not.

Have Win10pro and inside a VM with SL69.
The stalling is only in Windows-Boinc and not in SL69-Boinc on the same Computer (four Computer).

Could be the new theory version has triggered some server issues.

I hope that the problem with the extremely slow upload speed that many of your crunchers do see, but apparently not the staff of LHC@home, will come to an end but I have given up hope that it will happen anytime soon.
Compared with Rosetta that I am running now I can confirm that the upload speed to Rosetta in my case is 10 times higher\faster than it is on this project.

---

additionally, ATLAS result upload is still not working properly:

29.06.2018 08:28:40 | LHC@home | Starting task evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0
29.06.2018 10:47:45 | LHC@home | Computation for task evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0 finished
29.06.2018 10:47:47 | LHC@home | Started upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result
29.06.2018 10:50:04 | LHC@home | Backing off 00:03:16 on upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result
29.06.2018 10:53:22 | LHC@home | Started upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result
29.06.2018 11:02:01 | LHC@home | Backing off 00:07:07 on upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result
29.06.2018 11:09:09 | LHC@home | Started upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result
29.06.2018 11:11:23 | LHC@home | Backing off 00:12:22 on upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result
29.06.2018 11:23:46 | LHC@home | Started upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result
29.06.2018 11:26:35 | LHC@home | Backing off 00:19:15 on upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result
29.06.2018 11:47:19 | LHC@home | Started upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result
29.06.2018 11:49:47 | LHC@home | Backing off 00:40:19 on upload of evxMDm0ywssnyYickojUe11pABFKDmABFKDmZ0NNDmABFKDmt4oKNo_0_r1830430149_ATLAS_result

maybe the problem lies in the connection from inside CERN to outside CERN since you (the people inside CERN) dont see problems and we (outside CERN) experience problems.

A few notes on the download / upload problems:
- For me it seems that my Windows 10 host at home suffers more of this problem than my Windows 7 host. The win 10 machine has more CPUs and does a lot more work than the other
- The failing downloads (Atlas) never seem to go project backoffs or download pending but linger in state of "downloading" although no data is being transferred. Restarting the network connection gets data flowing again. This behaviour blocks also any sixtrack downloads when available.

---

Sorry, but as mentioned, on our BOINC clients upload speed is normal.

Well, can you just try to run boinc task from home?
On your own computer?

I'm now sitting with over 30 tasks that cant upload. Same machine is running other projects with no issues.

Have you tested:
https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4741&postid=35632#35632

I've tried disabling and re-enabling network activity

That gets the one or two stuck in download limbo kicked loose, but within one or two download files, the same behavior is observed

Downloads start normally,then "stall" where NO DATA is being transferred As a result the detected throughput drops to next t ozero

Have you tested:
https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4741&postid=35632#35632

Multiple times. It helps downloads but doesn't matter for uploads...they get 4 - 6 Mb uploaded and stall out. All my other projects have no issues on the same machine.

Have it too...
We hope they find a solution next week.

The situation didn't change significantly.

Result size: 123 MB

Di 03 Jul 2018 09:55:58 CEST | LHC@home | Started upload of nG4LDmA5FusnyYickojUe11pABFKDmABFKDmjs9aDmABFKDmsXWiim_0_r328315010_ATLAS_result
Di 03 Jul 2018 10:03:33 CEST | LHC@home | Backing off 00:02:54 on upload of nG4LDmA5FusnyYickojUe11pABFKDmABFKDmjs9aDmABFKDmsXWiim_0_r328315010_ATLAS_result
Di 03 Jul 2018 10:06:28 CEST | LHC@home | Started upload of nG4LDmA5FusnyYickojUe11pABFKDmABFKDmjs9aDmABFKDmsXWiim_0_r328315010_ATLAS_result
Di 03 Jul 2018 10:08:20 CEST | LHC@home | Finished upload of nG4LDmA5FusnyYickojUe11pABFKDmABFKDmjs9aDmABFKDmsXWiim_0_r328315010_ATLAS_result

Result size: 93 MB

Di 03 Jul 2018 10:11:32 CEST | LHC@home | Started upload of TlTLDmeXIusnlyackoJh5iwnABFKDmABFKDmb1qRDmABFKDmer21Lm_0_r312902500_ATLAS_result
Di 03 Jul 2018 10:28:17 CEST | LHC@home | Finished upload of TlTLDmeXIusnlyackoJh5iwnABFKDmABFKDmb1qRDmABFKDmer21Lm_0_r312902500_ATLAS_result

It's getting worse.

Result size: 107 MB
Upload speed: 16 kB/s

Di 03 Jul 2018 14:22:53 CEST | LHC@home | Started upload of G4CMDmIi3tsnlyackoJh5iwnABFKDmABFKDmsJMZDmABFKDmmCrP6m_3_r601011395_ATLAS_result
Di 03 Jul 2018 16:13:12 CEST | LHC@home | Finished upload of G4CMDmIi3tsnlyackoJh5iwnABFKDmABFKDmsJMZDmABFKDmmCrP6m_3_r601011395_ATLAS_result

I just abort them when I get a chance.

I saw these problems only with ATLAS in the past days. Tasks from all the other projects were downloaded with normal speed and without interruption.

My downloads are better...some still get stuck and apparently the "retry" option doesn't actual work...it just throws a "error while downloading". Uploads are starting to get through but still taking hours.