Development for LHC@home

Hi all,

If you don't use Linux you can ignore the rest of this post. If you do you may be interested in trying the experimental ATLAS app which doesn't use virtualbox but runs natively on Linux.

IMPORTANT!! To run this app you must install CVMFS, the CERN VM File System, and configure it for ATLAS. This file system contains all the software for ATLAS WU and is normally inside the virtual image (the same as for all LHC vbox apps).

A simple installation guide can be found here: https://cernvm.cern.ch/portal/filesystem/quickstart

You should set up the repositories as shown in the example for ATLAS. If you have a squid proxy handy you can specify it there - if not I'm not sure whether it will work or not without configuring one.

Our target for this app is CERN or ATLAS-related institutes who have idle machines with CVMFS already installed, and we do not expect the average volunteer to install CVMFS and run this app. But I think all of you here are above-average volunteers :) and you may be interested in trying it.

Please give feedback on the ATLAS forums. Unfortunately there is no way to check for CVMFS on the client before requesting tasks, so if you don't have CVMFS you can still get these tasks and they will fail straight away. So better to uncheck the ATLAS app if you don't want to run it.

Ubuntu 14.04, 16.04 - packages not found

sudo apt-get install cvmfs cvmfs-config-default
Reading package lists... Done
Building dependency tree
Reading state information... Done
E: Unable to locate package cvmfs
E: Unable to locate package cvmfs-config-default

[edit] instructions found

wget https://ecsft.cern.ch/dist/cvmfs/cvmfs-release/cvmfs-release-latest_all.deb
sudo dpkg -i cvmfs-release-latest_all.deb
rm -f cvmfs-release-latest_all.deb
sudo apt-get update

Unfortunately there is no way to check for CVMFS on the client before requesting tasks, so if you don't have CVMFS you can still get these tasks and they will fail straight away. So better to uncheck the ATLAS app if you don't want to run it.

without packages -

lhcathome-dev 23.02.2017 19:00:21 No tasks are available for ATLAS Simulation
lhcathome-dev 23.02.2017 19:00:21 VirtualBox is not installed

Unfortunately cvmfs is not yet in the standard Linux repos so you have to set up a separate repo.

without packages -

lhcathome-dev 23.02.2017 19:00:21 No tasks are available for ATLAS Simulation
lhcathome-dev 23.02.2017 19:00:21 VirtualBox is not installed

That is very strange, are you sure this message is for ATLAS? I have tested this app successfully on a machine without virtualbox.

That is very strange, are you sure this message is for ATLAS? I have tested this app successfully on a machine without virtualbox.

Yes, for the Atlas. Only one application is allowed in the settings.

Packages installed (including SQUID), configured and tested (CVMFS probe)
Error message remained the same (no tasks + no VirtualBox )

Now, however, they do not really have on the server. I am waiting for new work

Are there any additional requirements for the host for these tasks? The size of the memory or disk, the number of CPUs, etc.

This host has no virtualbox installed and runs ok.

But I see that I managed to drain the whole queue by myself overnight :) I have put more tasks in the queue and will put my machines on other projects to allow others to try this one.

The tasks are the same as the regular ATLAS@Home WU, so multicore (from 1 to 8 cores) and required memory is 2.3GB + 0.8 * ncores.

But I see that I managed to drain the whole queue by myself overnight :) I have put more tasks in the queue and will put my machines on other projects to allow others to try this one.

something drain queue a few times before you
but without having to install the necessary packages

it makes sense to further limit the number of tasks in one hand

BOINC9

lhcathome-dev 24.02.2017 11:00:37 Sending scheduler request: To fetch work.
lhcathome-dev 24.02.2017 11:00:37 Requesting new tasks for CPU
lhcathome-dev 24.02.2017 11:00:40 Scheduler request completed: got 0 new tasks
lhcathome-dev 24.02.2017 11:00:40 No tasks sent
lhcathome-dev 24.02.2017 11:00:40 No tasks are available for ATLAS Simulation
lhcathome-dev 24.02.2017 11:00:40 VirtualBox is not installed

do not understand why

That is very strange, are you sure this message is for ATLAS? I have tested this app successfully on a machine without virtualbox.

Yes, for the Atlas. Only one application is allowed in the settings.

Packages installed (including SQUID), configured and tested (CVMFS probe)
Error message remained the same (no tasks + no VirtualBox )

Now, however, they do not really have on the server. I am waiting for new work

Are there any additional requirements for the host for these tasks? The size of the memory or disk, the number of CPUs, etc.

what did you do for the http proxy?

i ran the probe and it failed.

for fun i tried to get tasks and i too get the virtualbox message

what did you do for the http proxy?

I don't have any proxy. So I setup a proxy directly to the host

sudo apt-get install squid

And added "http://127.0.0.1:3128" to config

Hello David,

Although not a real Linux native user, I gave it a try.

After some errors, the results show "Successfully finished the ATLAS job!" and are valid in BOINC.

But I'm not sure whether the results are useful. I set up tasks to use 2 cores/task.

Elapsed time ~10-12 minutes, but very low CPU. Maybe the low CPU is cause BOINC only counts the CPU-time of the wrapper and not of the other processes.

Host: https://lhcathomedev.cern.ch/lhcathome-dev/results.php?hostid=1778

Edit: Changed to single core tasks. The run time keeps the same.

specially installed virtualbox on the host
the problem remains - there is "no tasks for ATLAS"
it is necessary to look at the server - why tasks are not sent to the host

1 job finished successfully:
https://lhcathomedev.cern.ch/lhcathome-dev/result.php?resultid=311891

But:
- runtime was only a few minutes
- stderr.txt shows "mv: cannot stat `metadata-*.xml': No such file or directory ERROR: Missing metadata.xml"


Independent from this App:
The initial download per job is still more than 200 MB.
This is far too much as in case of an error every host pulls new jobs after only a few minutes and this would saturate nearly all kind of internet connections.

Between most valid tasks, I now also have some tasks with "Validate error",
although they run like the valid tasks and also have "Successfully finished the ATLAS job!" in the stderr.txt.

https://lhcathomedev.cern.ch/lhcathome-dev/results.php?hostid=1778

... I now also have some tasks with "Validate error",
although they run like the valid tasks and also have "Successfully finished the ATLAS job!" in the stderr.txt. ...

Same here.

Below some lines from .../BOINC/slots/x/pilotlog.txt

2017-02-24 13:44:34|16135|SiteInformat| !!WARNING!!2999!! $X509_CERT_DIR is not set and default location /etc/grid-security/certificates does not exist
2017-02-24 13:44:34|16135|aria2cSiteMo| WARNING: Rucio python modules not available
2017-02-24 13:44:35|16135|RunJobEvent.| RunJobEvent caught exception: No module named yampl
2017-02-24 13:44:35|16135|RunJobFactor| Failed to import RunJobArgo: Traceback (most recent call last):
File "/home/boinc/BOINC/slots/9/pilot3/RunJobFactory.py", line 36, in <module>
from RunJobArgo import RunJobArgo
File "/home/boinc/BOINC/slots/9/pilot3/RunJobArgo.py", line 16, in <module>
from MessageInterface import MessageInterface
File "/home/boinc/BOINC/slots/9/pilot3/MessageInterface.py", line 3, in <module>
import sys,os,ssl
File "/cvmfs/atlas.cern.ch/repo/sw/software/x86_64-slc6-gcc47-opt/19.2.4/sw/lcg/external/Python/2.7.3/x86_64-slc6-gcc47-opt/lib/python2.7/ssl.py", line 60, in <module>
import _ssl # if we can't import it, let the error propagate
ImportError: libssl.so.10: cannot open shared object file: No such file or directory

Do we need some configuration parameters that were previously unmentioned?

I am having problems

$ cvmfs_config probe
Probing /cvmfs/atlas.cern.ch... Failed!
Probing /cvmfs/atlas-condb.cern.ch... Failed!
Probing /cvmfs/grid.cern.ch... Failed!
$ ping atlas.cern.ch
ping: unknown host atlas.cern.ch
$ ping google.com
PING google.com (172.217.6.46) 56(84) bytes of data.
64 bytes from sfo03s08-in-f14.1e100.net (172.217.6.46): icmp_seq=1 ttl=53 time=19.2 ms

I can't reach the cern servers. Maybe it wants me to set a proxy? But I don't use a proxy.

---

Reno, NV
Team: SETI.USA

I can't reach the cern servers. Maybe it wants me to set a proxy? But I don't use a proxy.

cvmfs_config chksetup

if asks for proxy (CVMFS_HTTP_PROXY) - install squid
(Ubuntu - "sudo apt-get install squid")
and add CVMFS_HTTP_PROXY="http://127.0.0.1:3128"

you need proxies only on a single host
others can use it address

Although I recommend the use of a proxy you may try it direct.


Insert the following line in your /etc/cvmfs/default.local

CVMFS_HTTP_PROXY=DIRECT

then execute "cvmfs_config reload".

This is documented for Mac OS X but it may work on Linux also.

Although I recommend the use of a proxy you may try it direct.


Insert the following line in your /etc/cvmfs/default.local

CVMFS_HTTP_PROXY=DIRECT

then execute "cvmfs_config reload".

This is documented for Mac OS X but it may work on Linux also.

It worked! Thanks!

$ cvmfs_config probe
Probing /cvmfs/atlas.cern.ch... OK
Probing /cvmfs/atlas-condb.cern.ch... OK
Probing /cvmfs/grid.cern.ch... OK

---

Reno, NV
Team: SETI.USA

I am willing to give it a try, but what is the payoff?
That is, is there much of a performance advantage? If it works, will you be able to access it on the main LHC site?