R
RunPod•3mo ago
nerdylive

Why it seems like my job isn't assigned to a worker ( even after refreshing)

Read title
No description
No description
31 Replies
nerdylive
nerdyliveOP•3mo ago
I think it should be assigned after a few secs? this has been 10 seconds++ and it hasnt been assigned Also when i use this the other jobs doesn't get assigned to to the workers
Isaac
Isaac•3mo ago
I'm experiencing the same thing right now
nerdylive
nerdyliveOP•3mo ago
Oh on serverless? What were you running
NERDDISCO
NERDDISCO•3mo ago
so you are sending requests to your endpoint and those are not picked up by workers?
nerdylive
nerdyliveOP•3mo ago
Yes took too long, the job finnally got picked up like after 2mins~ this once
NERDDISCO
NERDDISCO•3mo ago
So is the situation better now? Or is it still very slow to pick up jobs?
nerdylive
nerdyliveOP•3mo ago
very slow, it should pickup faster This one request count scaling type should pickup jobs like in 4 secs or so right? but it took this "once" 111 secs ( from the delay time that i saw ) before it got assigned oh okay yeah it sometimes work it sometimes happens
Encyrption
Encyrption•3mo ago
What is the availability of the GPU type you picked in US-RO-1? Looking there now seems like your only good choices in US-RO-1 currently is 24GB GPU PRO and 48GB GPU PRO. The rest are unavailable or low availability. If you want to use another GPU I suggest you pick another region.
No description
nerdylive
nerdyliveOP•3mo ago
my workers are there i mean there are like 3+ workers idle
Encyrption
Encyrption•3mo ago
which GPU do you use?
nerdylive
nerdyliveOP•3mo ago
24 pro, 48 pro, 80gb its available my workers were idle
Encyrption
Encyrption•3mo ago
That should be fine, except for low availability for 80gb. It all depends on what is currently available in that region when you get a request. If there is not a GPU for you then it will just stay in QUEUE until there is. Next time you see your process stuck in QUEUE check the availability of your chosen GPU.
nerdylive
nerdyliveOP•3mo ago
.. yeah and thats not exactly my problem here so, for this case it hasn't been resolved yet for anyone just read
Encyrption
Encyrption•3mo ago
Sounds like a RP bug... looking at US-RO-1 it clearly seems to have the best availability for 24/48 pro.
nerdylive
nerdyliveOP•3mo ago
Okay it works now, nothing is wrong but runpod should look into this if they haven't made any changes just now haha
teddycatsdomino
teddycatsdomino•3mo ago
I'm running into something similar. I have a request in the queue that hasn't been picked up and a set of idle workers.
No description
No description
nerdylive
nerdyliveOP•3mo ago
Whats your scaling type?
teddycatsdomino
teddycatsdomino•3mo ago
No description
teddycatsdomino
teddycatsdomino•3mo ago
And you can see in the previous screenshot the request had been in queue for three minutes.
nerdylive
nerdyliveOP•3mo ago
Hmm yeah did this happens until now? Try to open support request on the website guys
teddycatsdomino
teddycatsdomino•3mo ago
I hadn't noticed anything like this until now, but I can't be certain it wasn't an issue before.
NERDDISCO
NERDDISCO•3mo ago
Do you still have this problem?
teddycatsdomino
teddycatsdomino•3mo ago
Yes, getting it right now.
No description
No description
teddycatsdomino
teddycatsdomino•3mo ago
No description
teddycatsdomino
teddycatsdomino•3mo ago
@Tim aka NERDDISCO Let me know if you need any other information.
nerdylive
nerdyliveOP•3mo ago
Yeah that scaling is weird, im using request count 1 for now ( quick fix )
teddycatsdomino
teddycatsdomino•3mo ago
For us that's unfortunately not a great option. We're processing video and our tasks run for 10-20 minutes. We don't want someone to be waiting for that long when there's an available worker.
nerdylive
nerdyliveOP•3mo ago
Yeah me too but it seems like setting scale type to req count 1 workd
NERDDISCO
NERDDISCO•3mo ago
@teddycatsdomino could you please use our official support mail at [email protected]? Because sadly I have no idea right now on how I can help you 😦
teddycatsdomino
teddycatsdomino•3mo ago
Of course. I honestly didn't even notice until @nerdylive posted about it! I may try his suggestion of setting the scaling type to "Request Count" with a maximum of one request to see if it does better.
NERDDISCO
NERDDISCO•3mo ago
@teddycatsdomino and if this is not working, please feel free to reach out. Because it might affect other users as well
Want results from more Discord servers?
Add your server