fah-client 8.4.9: HTTP_SERVICE_UNAVAILABLE ("No appropriate assignment")?

[calxalot]

Does restarting the client help?

After upgrading to macOS 15.6, I was getting repeated DNS failures. Not the same, I know.

[daiko]

The machines are running various versions of MacOS: 15.6, 13.7.7, and 26.0 (beta).
I restarted one, and while I was writing this reply, they have all picked up a WU. Coincidence? Perhaps there was about a two hour outage...fixed for now, it seems. Thanks, calxalot.

I was previously mostly getting WUs from the Tetrapeptide project; those seem to have disappeared. Now I’m getting T4 Lysozyme, Disordered Proteins, and AChE Inhibitors projects. Glad to be back in the fold.

[Schrödinger's cat]

Same here, it has been happening for over a week now.

Out of 12 GPUs, 7 have been retrying to download WUs for days with no luck.

Me too (1 of 1), but it's only been a couple of days.

[Schrödinger's cat]

Is there a reason GPU's can't do any WU?

Current jobs:

Code: Select all

Project Type   Public Jobs 
     0xa8        835,833
     0xa9          4,356
     0x22          1,695
     0x23              0
     0x24              0
     0x26             47
     0x27          9,611

There's a heap of FahCore_8 jobs, most likely due to toTow telling everyone you can't fold on a Raspberry Pi, unless it a 5B. My 5B's can do an 18240 WU in under 12 hours, with a Timeout of 35 hours. A 4B (perhaps half as fast) could complete do an 18240 WU before Timeout.

Anyway, why can't a GPU do one of these WU's. It would help reduce that 835,833 WU's to a better number (say, less then 50,000).

[muziqaz]

Some GPUs are too powerful for those projects.

[enroscado]

Some GPUs are too powerful for those projects.

Can you explain what do you mean by "too powerful"? Science wise it would make more sense to have GPUs doing "light" tasks than to have hundreds (or perhaps thousands) of GPUs idling. Wouldn't it?

[muziqaz]

Some GPUs are too powerful for those projects.

Can you explain what do you mean by "too powerful"? Science wise it would make more sense to have GPUs doing "light" tasks than to have hundreds (or perhaps thousands) of GPUs idling. Wouldn't it?

There are a large amount of people who moan about low PPD output when high end GPUs receive projects which are not well suited for them.
For those people loaded GPU which uses electricity, yet produces low PPD, is not worth the effort. Why do you think CPU folding is so unpopular? Because the PPD output from them is waaay smaller than with GPU.
If everyone thought like you, with priority to science, everyone would fold on CPU and GPU, instead of on GPU only. But they are not.
This shortage is temporary and things are improving.

[enroscado]

Did my question bother you? I am simply trying to understand.

I do not know why CPU folding is so unpopular, never really thought about it. Now that you mention it, a viable solution would be to increase the points awarded for CPU projects, that would get people folding on them (assuming those same WUs cannot be processed on GPUs).

Again, I was just trying to understand. Nothing else.

[muziqaz]

Did my question bother you? I am simply trying to understand.

I do not know why CPU folding is so unpopular, never really thought about it. Now that you mention it, a viable solution would be to increase the points awarded for CPU projects, that would get people folding on them (assuming those same WUs cannot be processed on GPUs).

Again, I was just trying to understand. Nothing else.

Where do you get the idea that your question bothered me?

[enroscado]

Where do you get the idea that your question bothered me?

Don't know, got that impression, Glad it didn't.

Thanks again.

[arisu]

Perhaps related to this assignment issue—All five of my Macs running CPU WUs are currently getting the same SERVICE_UNAVAILABLE message after successful uploads. This began about 90 minutes ago. All are set to use between 3 and 8 performance cores. Have the assignment servers run out of appropriate CPU WUs, or is there some problem with them? My network has no other problems that I can see.

I'm getting the same problem, which is surprising because CPU WUs are never in a shortage for anyone with >= 3 cores. After a few retries, it seems to get work. I've reported it in the internal development channel and it will be investigated.

One of the assignment servers is down (out of storage) with a separate error. I don't know if that's related. That is also being dealt with.

[arisu]

Some GPUs are too powerful for those projects.

Can you explain what do you mean by "too powerful"? Science wise it would make more sense to have GPUs doing "light" tasks than to have hundreds (or perhaps thousands) of GPUs idling. Wouldn't it?

I created a feature request to allow opting in to accepting light tasks during shortages: https://github.com/FoldingAtHome/fah-cl ... issues/375

Unfortunately the only reliable solution that doesn't come with downsides requires changes to the server code as well as the client code, and developer manpower is currently stretched pretty thin right now so it's not likely to be implemented soon. Perhaps an advanced option like "manually select GPU species" to override and spoof GPU would be viable instead. Any user that enables that would be on their own in terms of getting support, because it means that problematic WUs could no longer be blacklisted effectively on their hardware. But I don't think many users would enable that unless they had good reason.

[Schrödinger's cat]

For those of us who just want to help, could a GPU setting be set up in the 8.4 web interface to only download GPU WUs or download any WUs, with a note about reduced points.

Then, the servers could be set up to stop GPUs downloading any non-GPU WUs, if there are less than x of those WUs remaining.

[muziqaz]

For those of us who just want to help, could a GPU setting be set up in the 8.4 web interface to only download GPU WUs or download any WUs, with a note about reduced points.

Then, the servers could be set up to stop GPUs downloading any non-GPU WUs, if there are less than x of those WUs remaining.

You probably wrote in GitHub. I replied there that this feature will force a lot of rework, so don't expect it soon, if ever

[RamosXN40]

Perhaps related to this assignment issue—All five of my Macs running CPU WUs are currently getting the same SERVICE_UNAVAILABLE message after successful uploads. This began about 90 minutes ago. All are set to use between 3 and 8 performance cores. Have the assignment servers run out of appropriate CPU WUs, or is there some problem with them? My network has no other problems that I can see.

I'm getting the same problem, which is surprising because CPU WUs are never in a shortage for anyone with >= 3 cores. After a few retries, it seems to get work. I've reported it in the internal development channel and it will be investigated.

One of the assignment servers is down (out of storage) with a separate error. I don't know if that's related. That is also being dealt with.

I saw that when I was checking up on my raspi 4B (which DOES fold before timeout), I was going nuts trying to figure out what was out of space. Every time I checked df -h it said I was only using 12gb out of 128gb, and all the partitions were damn near relatively empty.

I know my other 2-cpu laptop will just sit there and wait for now, but everything was chugging along with my other computers until the raspi started complaining.


Update: the Mac and its 6 cores are down as well. Hopefully this just means it’s being worked on