Folding Forum

Hello,
I have two Server with 8CPUs both come with the error "No WUs available for this configuration" my computer at home with a "normal" i5 CPU is folding without interruption.
Is there something wrong with my configuration of these two Servers?

Would it be better to assign less CPUs and create more VMs?

Most likely this is just due to the load on the servers right now. I'll often have clients sit like this.

If you're really eager, pause and restart folding every 90 minutes to reset the retry timer. Don't do this every few minutes though, or you're only going to contribute to making the problem worse.

Worth a quick search of the forums before posting as there are tens (if not hundreds) of posts on varieties of this topic … but if you aren't sure then please do post … please also post logs so that assistance can be given ... it is possible you may have something that needs fixing but regrettably from the information provided the best answer anyone can given is that you are suffering as many are from the current issues with overloaded infrastructure caused by a massive recent increase in the folding community.

By saying you have a server with 8 CPUs, that is not too specific. How many CPU cores and threads does that represent?

For large counts of CPU cores/threads you may need to create multiple CPU folding slots with some fraction of the total assigned to each to get work. Usable CPU thread numbers will be multiples of 2 and/or 3. Multiples of 5 sometimes work, but not all projects will assign to those. Multiples of the prime numbers 7 and greater are not used.

At the moment there are a number of CPU projects getting released with a maximum of 12 threads. There are other projects out that can use more, not sure of the current max.

I am getting regular 30core slot WUs so there at least this large - just switching it to 32 to avoid the multiple of 5 so I can check that 32core slots are being fed.

Update ... There are definitely some WUs out there that are happy to be run on 32core slots.

I had similar issues on an iMac Pro with 20 cores.
I switched it to two slots, one 8 and one 12, downloading work isn't a problem anymore.
Maybe multiples of 4 are better? (Sorry I haven't searched the forum deeper yet)

=)

Dr._Rek wrote:I had similar issues on an iMac Pro with 20 cores.
I switched it to two slots, one 8 and one 12, downloading work isn't a problem anymore.
Maybe multiples of 4 are better? (Sorry I haven't searched the forum deeper yet)

=)

Still getting none available for this config sometimes, but getting assigned work more often. I switched to three slots of 8, 8, and 4 cores.
Might experiment with 5 slots of 4 cores later on.

Would normally be better for the science to run fewer higher count core slot (and better for points) avoiding the higher prime factors issue ... WUs get returned faster so the next gen WU can be generated and issued quicker ... this may also help reduce the overload of the servers a little bit ... so not sure what CPU count the OP has but hope some of this discussion has helped.

"No WUs available for this configuration"
means either
1) the FAH servers are too busy to handle your request
-or-
2) you need to change something in your configuration to get an assignment
-or-
3) the servers have run out of work for your perfectly configure system.

Right now, 1 is the most probable.
There's nothing you can do about 1 or 3 except to be patient.

With regard to 1: As to our efforts to add more server capacity, we're working as fast as we can.
With regard to 2, If you need help, we can help, but only if you post the key portions of FAH's log.
With regard to 3, you'll just have to wait for somebody who knows what research needs to be done next configures and releases a new project that does match your configuration. (Mostly based only on whether it's a CPU or a GPU. Each project only works on one, and not all GPUs are equivalent.)

Thank you for your answers.

I just wanted to know which configuration makes the most sense.
My host has 2 sockets with 14 cores each and I can create VMs with a maximum of 8vCPUs. So would it be better to create VMs with 8vCPUs or would it be better to create more VMs with less vCPUs?

This is because I noticed that my normal client is folding one WU after the other without any idle time and my VMs with 8vCPUs are waiting some hours to get the next WU

The team would probably prefer the fewer higher core VMs … Science gets processed quicker that way

There are a number of projects coming out for CPU folding that have a 12 thread maximum, so slots configured for 8 should start getting work. if the the between WU requests gets too long, more than a couple hours, you can pause the slot for a minute or two and then start it again. That resets the timer for waiting between requests. Please do not overuse this
, that can contribute to the network overload on the servers.

Joe_H wrote:... that can contribute to the network overload on the servers.

... and it doesn't help the COVAID research since the servers have to assume that any one client working on a WU is just as useful as another.

okay I understand.
Is there a way to see which projects with which configuration needs the most support. So I can configure the CPU slots with this number of CPUs needed?

As I understand it (and I may be wrong) it is not normally that a projects needs a specific number of cores - more of a case that some need less than a certain number of cores and other are probably best for larger core counts that can fold them quickly (very large atom counts or just "big" compute requirements) - most WUs just run on anything I believe … the allocation of WUs is down by Assignment Server which not hand out WUs to slots which don't meet minimum core counts, and as a rule will try not to issue a maximum core count WU to a larger core count slot (if it has no large one it can send a down step to the client and only use say 12cores of the 30cores in the slot (I had this recently) … It is not something that is configured from the client.

https://apps.foldingathome.org/psummary gives some information about the atom counts of projects … the amount of base points gives a fair indication of how much compute will be required … the timeout and deadline dates give an idea of the urgency of the project (in very crude terms) … so a large atom count with high basepoints and short timeout/deadline will probably not be folded quick enough by a low core count slot … not sure if they specifically identify the maximum core count projects here but keeping an eye on the beta team forum can give an indication when these are about to hit the main folding community.