Hello

I'm using BOINC 8.0.3 in Ubuntu 24.04. My issue is with trying to run more than 17 instances of Rosetta@HOME with 16 cores/32 threads (Ryzen 9 7950X3D) and 64 GB of RAM. I found that through enabling "mem_usage_debug" in the Event Log Options and viewing the event log, BOINC seems to be calculating or scheduling 3,500,000,000 Bytes exactly of RAM usage (Estimated WSS) for each Rosetta@Home instance. BUT, each instance only uses roughly between 200 and 500 MB (WS value from mem_usage debug). I don't know what WS or WSS stand for, so they might not be related, but check it out. I can run 7 instances on another machine that has 16 GB of RAM no problem. (BOINC 7.18.1, Ubuntu 22.04), so I don't think it's a Rosetta Issue.

The logs will probably explain what i'm trying to say better:
Both logs are taken soon after startup of boincmgr.
no mem_usage_debug: https://pastebin.com/QVJvLBxr
with mem_usage_debug: https://pastebin.com/fSxFyDET
my post on Rosetta@HOME with responses: https://boinc.bakerlab.org/rosetta/forum_thread.php?id=15377
^--please read that first, we went into some detail there.

Thanks,
Rich

Replied in your thread at Rosetta. Increase your pagefile size to 65GB.

Connected to this, 8.0.4 sorts out the problem of BOINC not recognising the statement in task files saying how much memory it might use. Older versions of BOINC would look at how much memory was being used by current tasks, not recognising that this might rise substantially. This means 8.0.4 will not overcommit on the memory front. I think there are only a handful of projects that this affects at the moment.

OK. I increased my swap file from 8GB to 64GB. I know you said 65, but if that's an issue, please tell me. lol.

Then I set my BOINC Manager to use 90% of swap instead of 80% to force an update (or so i thought). It's still only using 20 Rosetta@HOMEs (I know that's more, but it was doing that many before increasing swap, because some of them are Rosetta 4.20 instances, not 6.06... or at least i think that's why).

I went ahead and closed boincmgr and stopped and restarted boinc-client. With this much swap detected, shouldn't things get better? Nothing changed.

Here's my eventlog: https://pastebin.com/raw/uMw8cg8s

OK, uncheck any value in the Manager settings for storage allocation. Leave the host to be unlimited. I've seen the issue be solved this way also. The client doesn't/can't/gets confused sometimes interpreting which host allocation resource value is limiting a task and just reports one or the other values that can constrain a task from running.

Alrighty. I gave this a try. see screenshot:

After unchecking the boxes and saving. then exiting boinc manager, stopping the boinc-client service, starting the boinc-client service, and running boincmgr... Nothing seemed to change. I've got 16 Rosetta instances running and a GPUGRID instance running on my GPU. free -h says I'm using 9.6 Gi, with 44 Gi free.