Its been hot so Marks Rpi Cluster has been off. It cooled down yesterday so I have all the compute nodes running again. According to the weather forecast it should stay cool with showers for a week. If it gets too hot I will have to power down the cluster again.
The Pi3's are running Einstein@home BRP4 work. The Pi4's are running Einstein BRP4, Einstein Gamma Ray Pulsar search and some Rosetta@home work.
This is where Marks Rpi Cluster is at the end of December 2021 for Einstein@home. Total credit is 13,669,278 and recent average is 8,773 credits.
As you can see things were climbing quite nicely until the week before Christmas when it got hot and the cluster was powered off. Its now starting to climb back. Part of this would be effected by the cluster concentrating on Rosetta when it has work available.
The Pi4 compute nodes also do Rosetta@home in a 50/50 resource split. Rosetta don't have the graph like Einstein so I have shown the Certificate of Computation that they have. Total credit is 1,616,157 and recent average is 2,923 credits.
At the end of December 2021 the cluster is made up of:
3 x Pi4 (2GB) as support nodesThe Pi3's don't have enough memory (or CPU speed) so they do Einstein BRP4 work.
The Pi4's do Einstein BRP4 work, Einstein Gamma Ray Pulsar search and Rosetta@home work.
Today I took the opportunity to upgrade the last Pi in the cluster to Raspberry Pi OS based off Debian bullseye. Its a support node that most of the other nodes in the cluster use to connect to the internet which is why I didn't do it before. Its a Pi4 (2GB) with an external SSD plugged in via a USB3 port.
I took advantage of the Pi's ability to boot from USB devices this time, previously it was using a Micro SD card to boot. After flashing the SSD with a Raspberry Pi OS image I then plugged it into the Pi and off it went. No SD card needed any more. It shows up as device sda and the / and /boot partitions are on there. After that it was a matter of installing the software on it that it previously had, making sure the fstrim service was enabled and its done.
I'm feeling brave so might try and see if the old Pi Drives can also be used to boot off. I don't use them these days. I have a Pi as a storage server but its currently booting off an SD card. I wasn't sure I wanted to try imaging an external HDD in case it bricks it.
The cluster is still running 24/7. My RAC (recent average credit) for the Einstein project has been going up since running the Gamma Ray search tasks. Although they take much longer to run than the BRP4 tasks they give 10 times the credit. Rosetta has also started giving out work so after 3 weeks without any the Pi's that can run them are concentrating on it.
Update
A short while later and yes its possible to flash the Pi Drive and boot off it. It is a lot slower, although it might just be because its plugged into a Pi3 as well as being a slow hard disk.
And I can confirm the Seagate Expansion (desktop) HDD doesn't work. While you can image it the Pi won't boot to it and just hangs.
Marks Rpi Cluster continues running 24/7. The cluster is currently running Einstein@home work on all the compute nodes. Rosetta@home only had some python tasks which don't work on the Pi's.
The 5 x Pi3 nodes are running BRP4 work units (on 3 out of 4 CPU cores) at a time, so that gives 15 work units. The 9 x Pi4's are running a mix of BRP4 and Gamma ray work units on all 4 cores, giving another 36 work units in progress.
One of the Pi3's had a declining RAC (Recent Average Credit) compared to the other Pi's. Suspecting it was the micro SD card I re-imaged a new card and installed that last week. After a week of running the RAC is now climbing to be close to what the other Pi3's are getting. SD cards have a rudimentary wear-levelling mechanism so when its worn out it usually means all of the card is worn out. It wasn't giving any errors, just taking longer than the others to do the same thing.
