Headless Ubuntu Client won't continue without restarting service

[muziqaz]

Spoofing number of CPU cores will affect performance. Let's leave suggestion like that to ourselves, please.

[arisu]

Oh! I didn't mean to run more folding threads than there are hardware threads. Yes, that would pointlessly hurt performance. You'd have to turn the CPU slider down to match the real number of threads you have. Sorry for being unclear, never meant to suggest that!

[muziqaz]

Oh! I didn't mean to run more folding threads than there are hardware threads. Yes, that would pointlessly hurt performance. You'd have to turn the CPU slider down to match the real number of threads you have. Sorry for being unclear, never meant to suggest that!

But someone might interpret your suggestion as that

[arisu]

Very good point. And exactly why I didn't provide any guide.

[Knutowskie]

Well i got a free 4 core server with 24GB of RAM at oracles cloud. the cpu is not comparable to a RasPi.

We need to make clear, that "ARM is weak" belongs to the past. ARM is the future.

Oracle itself deletes unused VMs, so i gave it a use. Until i got time to do something better with it.

But lets NOT continue the discussion about cheating and so on. that is not the topic here.

all i wanted is: i dont want to MANUALLY restart the service until i get a WU downloaded. it works 100% when i restart the service. i get WUs. that is not the problem here.

[muziqaz]

Well i got a free 4 core server with 24GB of RAM at oracles cloud. the cpu is not comparable to a RasPi.

We need to make clear, that "ARM is weak" belongs to the past. ARM is the future.

Oracle itself deletes unused VMs, so i gave it a use. Until i got time to do something better with it.

But lets NOT continue the discussion about cheating and so on. that is not the topic here.

all i wanted is: i dont want to MANUALLY restart the service until i get a WU downloaded. it works 100% when i restart the service. i get WUs. that is not the problem here.

We need to be clear that people need to get off the ARM hype train. ARM is the future message has been the thing for more than a decade. Still see no good desktop platform to play with as an alternative to x86 desktop. ARM is the future for large corpos who built CPUs for their own good (apple, among others). For general public, there is absolutely no reason to even dream of any alternatives to high performance x86

[calxalot]

Please set CPUs to 4 and retry.

[foxpy]

Well i got a free 4 core server with 24GB of RAM at oracles cloud. the cpu is not comparable to a RasPi.

We need to make clear, that "ARM is weak" belongs to the past. ARM is the future.

Oracle itself deletes unused VMs, so i gave it a use. Until i got time to do something better with it.

But lets NOT continue the discussion about cheating and so on. that is not the topic here.

all i wanted is: i dont want to MANUALLY restart the service until i get a WU downloaded. it works 100% when i restart the service. i get WUs. that is not the problem here.

Actually, before you begin folding on this server, are you sure that this free tier VM has 4 dedicated CPU cores, or are they simply the shared ones? The performance of shared vCPUs is going to be extremely bad for fah due to GROMACS nature of running as fast as the slowest thread.

[arisu]

Well i got a free 4 core server with 24GB of RAM at oracles cloud. the cpu is not comparable to a RasPi.

We need to make clear, that "ARM is weak" belongs to the past. ARM is the future.

Oracle itself deletes unused VMs, so i gave it a use. Until i got time to do something better with it.

But lets NOT continue the discussion about cheating and so on. that is not the topic here.

all i wanted is: i dont want to MANUALLY restart the service until i get a WU downloaded. it works 100% when i restart the service. i get WUs. that is not the problem here.

Actually, before you begin folding on this server, are you sure that this free tier VM has 4 dedicated CPU cores, or are they simply the shared ones? The performance of shared vCPUs is going to be extremely bad for fah due to GROMACS nature of running as fast as the slowest thread.

If OP is using the Oracle Always Free tier, then that means he's using OCI Ampere A1 Compute which gives 4 "OCPUs" which is Oracle's name for a type of virtual CPU. I don't know if each one runs at full speed or if each one is already shared with other tenants, but more is always better and each additional one you get adds a single OCPU worth of processing power. I think that each one is guaranteed to give you a certain performance and you won't have to suffer from 3 fast ones and 1 slow one that happens to be getting used by someone else.

To OP, if you are using the Always Free tier, you'll have to keep network utilization above 20% to not get it shut down, not just CPU utilization via FAH.

[foxpy]

I think that each one is guaranteed to give you a certain performance and you won't have to suffer from 3 fast ones and 1 slow one that happens to be getting used by someone else.

Well, as long as this setup finishes work before timeout, I think it is worth folding on, especially for such a low price

[Knutowskie]

I think that each one is guaranteed to give you a certain performance and you won't have to suffer from 3 fast ones and 1 slow one that happens to be getting used by someone else.

Well, as long as this setup finishes work before timeout, I think it is worth folding on, especially for such a low price

The low price of 0€ is best about it.

But still i am stuck on DOWNLOADING forever until i restart the service...

Code: Select all

fahctl state
{
  "info": {
    "version": "8.4.9",
    "url": "https://v8-4.foldingathome.org/",
    "os": "linux",
    "os_version": "6.8",
    "cpu": "arm64",
    "cpu_brand": "Neoverse N",
    "cpus": 4,
    "gpus": {},
    "mach_name": "KNUT-02",
    "hostname": "KNUT-02",
    "mode": "Release",
    "revision": "360fe71b1bd05bb89814bfb97b73a5bda84802d6",
    "has_battery": false,
    "on_battery": false,
    "id": "################################",
    "account": "##############################"
  },
  "config": {"user": "Knutowskie", "team": 70335, "passkey": "#########################", "cause": "unspecified"},
  "groups": {
    "": {
      "config": {
        "on_idle": false,
        "on_battery": false,
        "keep_awake": false,
        "paused": false,
        "finish": false,
        "beta": false,
        "cuda": true,
        "key": 0,
        "cpus": 4,
        "gpus": {}
      },
      "wait": "2025-04-05T21:28:27Z",
      "failed_wus": 0,
      "lost_wus": 0,
      "failed": ""
    }
  },
  "units": [
    {
      "group": "",
      "number": 34,
      "cpus": 4,
      "gpus": [],
      "state": "DOWNLOAD",
      "paused": false,
      "id": "VR0kBC3fX_aN_UYszQGbWMMzWbfbwhUMjNGhcldhMbI"
    }
  ]
}

so my "headless" machine is not headless at all. thats the point here.

Code: Select all

00:01:08:I1:WU15:Requesting WU assignment for user Knutowskie team 70335
00:01:08:I1:OUT1:> GET https://api.foldingathome.org/machine/################## HTTP/1.1
00:01:08:I1:OUT2:> GET https://api.foldingathome.org/gpus HTTP/1.1
00:01:08:I1:OUT3:> POST https://assign1.foldingathome.org/api/assign HTTP/1.1
00:01:09:I1:OUT1:< HTTP/1.1 200 HTTP_OK
00:01:09:I1:OUT4:> GET wss://node1.foldingathome.org/ws/client HTTP/1.1
00:01:09:I1:OUT2:< HTTP/1.1 200 HTTP_OK
00:01:09:I1:OUT4:< HTTP/1.1 101 HTTP_SWITCHING_PROTOCOLS
00:01:09:I1:Logging into node account
00:02:03:I1:OUT3:< HTTP/1.1 200 HTTP_OK
00:02:03:I1:WU15:Received WU assignment B1cvcGLXQa2NNEpvOw8zCm9cPwrN_NeOpHOR8JRnXJg
00:02:03:I1:WU15:Downloading WU
00:02:03:I1:OUT6:> POST https://highland2.seas.upenn.edu/api/assign HTTP/1.1
00:07:42:I1:Clean exit

Maybe we can focus on the problem here and not go into theoretics about OCPUs or ARMvsx86

I always get it to work when i restart the service. So basically something is preventing it from finishing the download or something and the logs dont tell me why.

[muziqaz]

many layers of cloud computing? Many things might be preventing to download WUs. When you have a computer which is at your house/flat and it is your computer, you can troubleshoot, because you are the one who set up the network and computer itself
When you have cloud instance, none of it is your doing, and all of it is behind firewalls and other bureaucratic stuff.

[foxpy]

To be honest for me this looks like some bug in fah-client. Maybe it lacks a timeout somewhere and tries to read from a stuck socket indefinitely. If I were you, I would report it on github here: https://github.com/FoldingAtHome/fah-cl ... tet/issues

[arisu]

To be honest for me this looks like some bug in fah-client. Maybe it lacks a timeout somewhere and tries to read from a stuck socket indefinitely. If I were you, I would report it on github here: https://github.com/FoldingAtHome/fah-cl ... tet/issues

Yes it is a bug. It can sometimes get stuck in an uploading state too, which in some cases triggers a secondary bug that puts the database into an inconsistent (but harmless) state: viewtopic.php?p=368629#p368629