27th of August

Marks Rpi Cluster continues to run 24/7. They are all running Einstein@home BRP4 work at the moment.

As I mentioned in my last post BOINC has a concept of a back-up project. For the 8 Pi's that had Asteroids@home I set the resource share to 0% which is how you configure it.

I ordered 20 more micro-SD cards in anticipation of an updated Raspberry Pi OS. There was some discussion in the Raspberry Pi forums that there is a closed beta-test happening. It was also mentioned in a different message thread that they may want to upgrade to the 6.4 or 6.5 kernel as more Pi-specific modifications have been up-streamed into the standard kernel.

My 4th Pi4 8GB reached 1M credits for Einstein this weekend. This particular Pi joined the project on the 3rd of April 2021, the same time as the other three Pis that have already reached 1.2M.

20th of August

Marks Rpi Cluster continues to run 24/7.

The day after my last post Einstein@home had another outage for 2 days. Unlike last time this outage was planned, although I didn't see the notice until afterwards. They use the Atlas cluster in Germany and were doing maintenance on their power infrastructure. As I did last time I resumed running Asteroids@home work.

BOINC has a concept of a backup project, that is one that does work when you can't get it from other projects. I haven't got my Pi's configured this way. Instead I need to specifically tell them to allow work from Asteroids. I might change Asteroids to be a backup project so I don't need to intervene when Einstein runs out of work.

I mentioned in my last post about replacing the MicroSD cards in all the compute nodes. I just had a look through my SD card spares box and I only have 6 spares (of the 32GB high endurance variety) so I need to order more. I went onto the JB Hi Fi website and it seems it will only let me order 5. Stupid website. I suspect it is checking the local store stock level and that is probably all they've got. I did try changing to another store but it didn't help. It looks like I will have to go in-store to order them.

13th of August

Marks Rpi Cluster continues to run 24/7. Since my last post Einstein@home has run out of the BRP4 work twice. Once I notice I usually allow Asteroids@home work on those Pi's that have the project (which is 8 out of 24 Pis). Their work outages usually last a week or so before things return to normal. From what has been said on their forums the pipeline to create the work units needs regular monitoring and fixes.

I think the SD cards are close to wearing out so I expect I will need to replace all of them. These days I have been buying the 32GB Sandisk high endurance cards, but don't have enough spares on hand to replace all of them. I was hoping the Raspberry Pi foundation would have the latest OS based upon Debian bookworm out as I would need to reinstall all the Pi's for it anyway. If I can do that while replacing the SD cards it would save time.

Talking about SD cards I try to minimize the writes to make them last longer. I have BOINC set to checkpoint every 5 minutes and keep apps in memory (as the compute nodes are Pi4 8GB). I also have Linux set to limit the log files to 100MB.