I had another charger die. The strange thing about this one was I had 4 Pi3B+ spread across two chargers, so it was only powering two Pi's. That is well within its rated capacity. That puts the cluster down by 2 Pi3B+ nodes at the moment.
It comes up regularly what people use to power multiple Pis and the most common answer seems to be either make your own power supply or use an Anker USB charger. I bought an Anker 5 port USB charger two years ago but the Pi's wouldn't even power up with it which is why I am using the ones I have now. The Anker works great for charging the phone or tablet, just not for running a Raspberry Pi.
A week ago the official Pi4 cases arrived. The case, power pack and Pi4's were all ordered together from the same supplier. The case comes in 2 parts. They have no air holes, just cut-outs for the connectors and that is it. They are totally unsuitable. I was thinking I would cut a hole in the top and mount a 40mm fan however the air has to go somewhere and that is a problem with the new case design.
This week the Pi4's arrived. Unfortunately the power packs are still on back order so I can't use them yet. Also the 40mm 5v Noctua fans arrived. At this point I can't do anything until the power packs arrive so while I am waiting for them I will look into what I can do with cases and fans.
Sunday, August 25, 2019
Saturday, August 3, 2019
Increasing the bramble
I decided to increase the size of the cluster - just at the wrong time.
I have a number of older Pi's (4 x Pi2B and 6 x Pi3B's) sitting around gathering dust. I have enough power adapters and I managed to find enough network cables to get them going. I even found a spare 6 socket power board. After a bit of furniture rearranging I had somewhere to put them. I flash a bunch of SD cards for them and bring them up one by one. Everything is going well so far.
Then I find that a number of the existing compute nodes only have a couple of tasks instead of the usual 4 each. The new ones aren't getting work either. After a bit of looking at logs it seems Einstein's scheduler is telling me it hasn't got any work. A quick check of the Einstein server status page tells me otherwise. I have a look at their scheduler log and it doesn't offer any clue other than it doesn't think there is any work available.
That brings Marks Rpi Cluster up to 12 Pi3B+ nodes and 10 Pi3B as compute nodes. I didn't use the Pi2B's as I think they are too slow and besides I don't have enough power adapters or network cables.
While I was at it I also updated the nfs server (a Pi3B+ with a PiDrive) to Raspbian Buster. I had to find my original blog post on how to setup an nfs server. It was the last Pi that I needed to update so now the whole cluster is running buster.
I have a number of older Pi's (4 x Pi2B and 6 x Pi3B's) sitting around gathering dust. I have enough power adapters and I managed to find enough network cables to get them going. I even found a spare 6 socket power board. After a bit of furniture rearranging I had somewhere to put them. I flash a bunch of SD cards for them and bring them up one by one. Everything is going well so far.
Then I find that a number of the existing compute nodes only have a couple of tasks instead of the usual 4 each. The new ones aren't getting work either. After a bit of looking at logs it seems Einstein's scheduler is telling me it hasn't got any work. A quick check of the Einstein server status page tells me otherwise. I have a look at their scheduler log and it doesn't offer any clue other than it doesn't think there is any work available.
That brings Marks Rpi Cluster up to 12 Pi3B+ nodes and 10 Pi3B as compute nodes. I didn't use the Pi2B's as I think they are too slow and besides I don't have enough power adapters or network cables.
While I was at it I also updated the nfs server (a Pi3B+ with a PiDrive) to Raspbian Buster. I had to find my original blog post on how to setup an nfs server. It was the last Pi that I needed to update so now the whole cluster is running buster.
Subscribe to:
Posts (Atom)