Rosetta (Folding) at home requires twice the available memory following an update.

The Rosetta@home project stopped providing me with fresh work units yesterday, apparently due to a server issue their end. My 4GB Pi 4 ran out of work units overnight and did not retrieve any more so I took the opportunity to reboot my Pi and check for updates.

It would appear that my Pi did update its Balena client and following on from that it is not retreiving any more work units. Rosetta is showing units as being ready but the GUI on my Pi is showing the following after each scheduled update:

“Message from server: Rosetta needs 6675.72MB RAM but only 3653.61MB is available for use.”

Can we confirm that Balena has not introduced this error ? I can see the “server” being at both the client end and that at Rosetta.

My current build is f79ec6d (balenaOS 2.73.1+rev1).

Update - while typing this my Pi obtained more work units but the error message relating to insuffucient memory remains.

This seems to have been an issue at the Rosetta end - a number of other users have subsequently complained about unusually large work units being passed across. Since I raised this issue here I have been able to download a fresh cache of work units but a similar number were aborted during the download suggesting that Rosetta are still having issues of their own.

The Rosetta forum is still getting posts from people complaining about the sudden increase in system requirements. It is increasingly looking like the minimum memory requirement for a Pi is now 8GB with lower spec variants no longer receiving work units. Booo !

Same situation here, but is a general scarcity of works. The rosetta server activity ‘tasks in progress’ has dropped by around 200,000 (a third).
https://munin.kiska.pw/munin/Munin-Node/Munin-Node/results_rosetta.html

So just wait for a future server-side solution.

Thank you for the confirmation. I was until fairly recently monitoring the Rosetta stats and forums but I got the impression that the latter is not always the most friendly place to be unless you’ve got a 5 figure average credit and thousands of cores on the go. I might give it a couple of months and try it again but for now the Pi has been repurposed elsewhere.

I’ve been in the same position. I have 2 Raspberry Pi 4 and one Generic AARCH64 board. I made around 1200 credits daily, not much but running constantly for a year I’m going to reach a decent figure when the year is complete. But I think the Boinc community must think the third part of the works are gone when they left behind all the small boards.

It’s good to hear from someone else volunteering “slow and steady” hardware to the project.

Someone on the Rosetta forum also mentioned about a third of the work-force possibly becoming obsolete and no-one from the project picked up on it. There was also some speculation that the issue might have been some badly parsed work units claiming a greater demand on resources than they were ever likely to require but no-one who answered me seemed to grasp that these were choking the lower end hardware through aborted or rejected downloads.

The funny thing about the Rosetta forums is how readily they will defend the volunteers who run the project while berating the volunteers who donate their resources and pay to run and maintain their own kit. I know my single Pi will never set the world on fire but it’s still costing me to run it and I do not think it unreasonable to be able to cache at least a day or two’s worth of work units without being told off by the forum members for hoarding. In the relatively short time I had been contributing the project my little £40 SBC running off an SD card had longer uptime than the Rosetta servers and I was made out to be the bad man by people running monster rigs capable of processing thousands of tasks a day for questioning the management of the project.

I might look into other BOINC projects at some point if I can work out how to get them working on a Pi (I’m still learning this stuff) but the attitude of some of their forum members and the lack of response from Rosetta itself (volunteers or not) has really taken the shine off of my participation.

1 Like

I try to explorer other projects supposedly have ARM task. But I can’t find the way to make it work in th Pi. See this thread with my test:

I try the past week again with the same results. No tasks for my Pi

I believe that the Einstein project was one of those that I had looked at previously as well. I would like to direct you to an earlier post where one of the Balena team answered a question I had regarding unsupported projects listed in the Folding client here:

… in particular @jtonello mentioned that Berkeley had made a custom image for Balena to use with Rosetta. It still seems to be the case (ie a quick Google on a mobile phone) that the Pi does not have much love or support in the realms of distributed computing because they cannot complete (large) work units fast enough. I would like to see Balena making more projects easily accessible but it clearly requires close co-operation with third parties to make it work, and given the largely volunteer driven eco-system I do not want to make Balena out to be lacking in any way here.

Personally I think small but high up-time SBC’s and the like are a wasted resource since they are cheap and out there in large numbers often owned by enthusiasts and tech savvy hobbyists. My other regular hardware platform is an i7 based laptop that would clearly be “better” than a Pi for knocking out work units and more widely supported than ARM. However it’s only on for a couple of hours a week and therefore would not complete any work units within their allotted time. But I do have Pi’s on 24/7, including one that was dedicated solely to the task thanks to the Balena image. It might have been comparitively slow but it was reliable and consistent.

Thinking about my home at the moment at any given time I have at least 3 Pi’s permanently connected and doing things with only one of them under any real load. The other two barely register any load at all over a 24hr period and would be ideal candidates for running distributed computing on in their spare time since they have lots of that. I will look into running distributed projects on the Pi again when I have the time and if I find anything that looks particularly interesting I will post it back here.

1 Like