Computers not found / hardly any tasks

Message boards : Number crunching : Computers not found / hardly any tasks

To post messages, you must log in.

AuthorMessage
fritz

Send message
Joined: 17 Apr 20
Posts: 3
Credit: 1,626,120
RAC: 0
Message 94836 - Posted: 19 Apr 2020, 7:32:48 UTC

Hi,
I'm new here and my company wants to dedicate a few hundred cores to this projects. We have hundreds of RPI3B+ in stock and until we ship them to our customers, we want to let them run rosetta@home.

We're using the https://foldforcovid.io/ image, provided by balena.io

Until now, I've set up 13 RPIs. but on the "Your computers" page, only 3 of them are visible:
https://boinc.bakerlab.org/rosetta/hosts_user.php?sort=rpc_time&rev=0&show_all=1&userid=2143308

I assume, the computer infomation can be found in the global_prefs_override.xml file. If I look into this, I see e.g. <host_cpid> and <externl_cpid> -> I assume, this would be the computer Id?
Interestingly, some share the very same Ids, some don't, why is that?

I just added the weak account project id to all, so they are all gathered together into one account / team.

Also, just 2 or so ever received any computation tasks and just one was accounted.

Would be very glad if someone finds time to help me out setting this up correctly, so we can increase the fleet next week!

Best regards
Fritz
ID: 94836 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1677
Credit: 17,761,983
RAC: 22,819
Message 94841 - Posted: 19 Apr 2020, 7:58:41 UTC
Last modified: 19 Apr 2020, 8:00:37 UTC

This sounds very much like a problem someone had over at Seti with a whole bunch of Pis they wanted to use, and it appeared they were all processing work, but only a few of them were showing up on their Account.
It doesn't look as though the issue was ever resolved, but if you checkout the thread, something there may be of use.

When you attach a system to BOINC, it should get it's very own ID number. There are cases where if there are communication issues between a system and the servers a system can end up with a new ID number. But how you can have a bunch of systems, and only a few of them show up, i've no idea. And i've no idea how some system could get the same IDs.
And it appears to be an issue with Pis only. Many others in the past have had setups of dozens (even hundreds) of systems, on their account without this problem occurring.
Good luck.

Multiple computers setup?
Grant
Darwin NT
ID: 94841 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 94878 - Posted: 19 Apr 2020, 14:30:02 UTC - in response to Message 94841.  

Is the uniqueness of the host lost when you duplicate a working setup from one to another? Does the server actually think more than one of them are the same system? An audit of the active WUs ought to prove if something like that were the case.
Rosetta Moderator: Mod.Sense
ID: 94878 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Bryn Mawr

Send message
Joined: 26 Dec 18
Posts: 390
Credit: 12,073,013
RAC: 4,827
Message 94909 - Posted: 19 Apr 2020, 17:15:50 UTC - in response to Message 94878.  

Is the uniqueness of the host lost when you duplicate a working setup from one to another? Does the server actually think more than one of them are the same system? An audit of the active WUs ought to prove if something like that were the case.


In which case would changing the system name separate them in the eyes of the server?
ID: 94909 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
fritz

Send message
Joined: 17 Apr 20
Posts: 3
Credit: 1,626,120
RAC: 0
Message 94921 - Posted: 19 Apr 2020, 18:10:41 UTC - in response to Message 94878.  

Hi,
the balena concept allows the duplication of the base system. But every system (on first boot) generates its unique hostname and can be maintained from a web UI dashboard.
(think of it as a raspian image -> everyone downloads the same, but once installed, each system is unique).
In short, I did not duplicate a working system, I just used the same base image to deploy it to a number of RPIs.

Thanks
ID: 94921 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1677
Credit: 17,761,983
RAC: 22,819
Message 94947 - Posted: 19 Apr 2020, 22:53:13 UTC - in response to Message 94921.  

Hi,
the balena concept allows the duplication of the base system. But every system (on first boot) generates its unique hostname and can be maintained from a web UI dashboard.
(think of it as a raspian image -> everyone downloads the same, but once installed, each system is unique).
In short, I did not duplicate a working system, I just used the same base image to deploy it to a number of RPIs.
How did you install BOINC on those systems and then attach them to projects?
Grant
Darwin NT
ID: 94947 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
fritz

Send message
Joined: 17 Apr 20
Posts: 3
Credit: 1,626,120
RAC: 0
Message 94968 - Posted: 20 Apr 2020, 6:19:43 UTC - in response to Message 94947.  

Hi Grant,

I've used the setup here:

https://github.com/balenalabs/rosetta-at-home

Simplified, it is a docker container with a custom install script for the boinc client.
For the RPI (because of the 1GB RAM limit, it is started then as:
boinc --allow_remote_gui_rpc --fetch_minimal_work


Interestingly, yesterday evening more computers started to get work.
Now I have 6 listed, whereas 2 of them still have 0 credit, one is still with total credit of 25 (which is now the 3rd day without change), and the other 3 have total credits of: 4, 135 and 242. Very strange distribution.

Cheers
Fritz
ID: 94968 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : Computers not found / hardly any tasks



©2024 University of Washington
https://www.bakerlab.org