its GCP..I spin up a box through auto scaller

I cloned one of my task which was successfully completed, and I am trying to run it through the autoscaler by enqueing it. No idea why its been crashing again and again,

isn't it strange that one of them working and other got failed? and also when the config says 5 instances why it spun up 8 instances? Any idea about it

I remove credentials from the config for security purpose, and replaced it with XYZ

{
    "gcp_project_id": "XYZ",
    "gcp_zone": "us-central1-b",
    "gcp_credentials": "XYZ",
    "git_user": "mkerrig",
    "git_pass": "XYZ",
    "default_docker_image": "pytorch/pytorch:1.7.0-cuda11.0-cudnn8-runtime",
    "instance_queue_list": [
        {
            "resource_name": "v100",
            "machine_type": "n1-highmem-4",
            "cpu_only": false,
            "gpu_type": "nvidia-tes...

and also 4 of my task failed but the 5th one runs completely fine,

it's odd to me as well, what's strange is that, the error was messaged was related to Nvidia and that wasn't the case as well. Anywho, its solved I am happy!!

I am on GCP rn, not working with AWS

I can see my process being crashed, I need to know why..if this is the clone of a completed task, then why it produces such error?

it spins up 8 instances even though my config says 5

what can be the cause of this

I enqueued 5 task to this auto scaler, 4 of them failed but the 5th one is working as expected..those enqued task are the clone of completed task

same config no change.

Hey, whats the rate limit?

Got it.

strangely, I reset my task and qued them again, and its all working..

@<1571670393394040832:profile|LittleDolphin60>

I wanna know the language..

bash: line 1: 1031 Aborted (core dumped) NVIDIA_VISIBLE_DEVICES=all $LOCAL_PYTHON -u -m clearml_agent execute --disable-monitoring --id b7af4abc3c174a32a5612c5c27a7b523
2023-05-24 17:10:00
Process failed, exit code 134

I used the same config 1 month ago, but its not working today, I have run 100s of experiment on the auto scaler, but its not working now, no idea why.

@<1571670393394040832:profile|LittleDolphin60>

ideally, how much should I wait, once I put my task in some que?

2023-05-22 11:14:25,238 - clearml.Auto-Scaler - INFO - --- Cloud instances (8) ---
2023-05-22 11:14:25,238 - clearml.Auto-Scaler - INFO - 1110204948425405426, regular
2023-05-22 11:14:25,238 - clearml.Auto-Scaler - INFO - 1364006518840029853, regular
2023-05-22 11:14:25,238 - clearml.Auto-Scaler - INFO - 4551653386764087872, regular
2023-05-22 11:14:25,238 - clearml.Auto-Scaler - INFO - 4704932875408438200, regular
2023-05-22 11:14:25,238 - clearml.Auto-Scaler - INFO - 48755565930452715...

Sorry dear, needed urgent help.

Any who I solved by commenting my code and created bunch of clone task and uncomment the code line by line, the issue was related to task.connect being called multiple times I guess.

I am such a newbie, Sorry again.