I recently ran into an issue where Nomad has reserved too much CPU & memory althoough configured otherwise. This can result in tasks being killed by the OOMKill.

1OOM Killed

Workaround

There is a workaround by overriding the available memory and/or CPU power in your Nomad client config file.

Nomad client config

You can compute your "real" CPU compute power by multiplying the number of cores with the MHz. Example: 4 core CPU with 2 GHz each = 4 * 2000 = 8000.

To get your CPUs clock speed, use the lscpu util:

1lscpu | grep MHz

The Nomad client config is located at /etc/nomad/nomad.hcl by default.

1client {
2  memory_total_mb = 16000
3  cpu_total_compute = 8000
4}

Sources

• client Block - Nomad Docs

• Allow for 0 res allocations / Disable resource constraints for particular agent #5672

• Hashicorp Nomad and the memory options with the docker driver

Read more...

No, you can't use your $6,299.00 Camera as a Webcam. That will be $5

$30 billion company want's you to pay $5 for using your own camera as a webcam

17 Jan 2025

|

4 min read

Fix Laravel Job Queue not processing with MaxAttemptsExceededException

Fix issues with the Laravel Job Queue not processing jobs due to max attempts exceeding.

12 Jan 2025

|

2 min read