The direction that we are going to take with this is as follows:

- For workunits that have copies already sent out, we are changing the resource configuration for memory to reflect the much larger RAM used (either 4GB or 1.5GB depending on the specific job). We are also boosting the disk size to accommodate extra size. This will allow BOINC to manage sending resends to only computers that can handle them (it will also keep multiple copies from running at once)
- For the other jobs we are going to rebuild them but limit them to 20 per workunit so that this issue doesn't consume massive memory. This will result in a period of quick jobs with a little bit larger than normal memory footprint, but it cause the issues like these have.
- Longer term the memory leak will need to be fixed (it is 14 MB per job within the workunit)

Testing the changes and getting things in place will take a bit time so CPU work will be stopped for the next 24-48 hours while we prepare and test.

The kraken's have been released.


We have re-enabled sending worker for OpenPandemics CPU. However, at this time the only jobs available are the repair jobs for the workunits that require large memory. The workunits have been modified to state that they need either 4 GB or 1.5 GB depending on the job. This will ensure that they run successfully.

There are only 48,000 of these at the moment, but given the high requirements I don't know how long it will take to distribute them.

We continue to work on recreating the other batches with limits to prevent the high memory use.

I thought we were talking about more GPU WUs.

Actually - that is exactly what we want to have happen. When I wrote



What this means is that the memory limit for these jobs has been reset so that those jobs will not be sent to computers with insufficient memory to handle them. Your raspberry pi is protected from these jobs because of this limit.

Most of the 4GB resends have now been sent out and we are about to start going through the jobs that require 1.5 GB. There are currently 21,888 of them to send out. Once we get through those we will then be sending out much more normal sized jobs and your computer should start to receive them at this time. That should be in another hour or two.