OOM using mapt 0.12.1

[albfan]

Using task on a pipeline it is killed due to OOM:

        taskRef:
          params:
            - name: url
              value: 'https://github.com/crc-org/ci-definitions'
            - name: revision
              value: crc-builder-v1.1.2
            - name: pathInRepo
              value: crc-builder/tkn/crc-builder-arm64.yaml
          resolver: git
        status:
          artifacts: {}
          completionTime: '2026-06-17T14:38:03Z'
          conditions:
            - lastTransitionTime: '2026-06-17T14:38:03Z'
              message: '"step-provisioner" exited with code 137: OOMKilled'
              reason: Failed
              status: 'False'
              type: Succeeded
          podName: nightly-microshift-linux-arm64-ivpxiu-build-arm64-binary-pod

Seing #175, should it increase again cpu and memory?

Running locally (i7, 64Gb) there's no issue

Maybe related with lot of errors like:

�[31mERRO�[0m GET https://management.azure.com/subscriptions/b0ad4737-8299-4c0a-9dd5-959cbcf8d81c/providers/Microsoft.Compute/locations/malaysiawest/communityGalleries/Fedora-5e266ba4-2250-406d-adad-5d73860d958f/images/Fedora-Cloud-42-Arm64
--------------------------------------------------------------------------------
RESPONSE 404: 404 Not Found
ERROR CODE: NotFound
--------------------------------------------------------------------------------
{
  "error": {
    "code": "NotFound",
    "message": "please check if the gallery name 'Fedora-5e266ba4-2250-406d-adad-5d73860d958f' is valid and the gallery exists in 'MalaysiaWest' region."
  }
}

I can try to increase the memory:

https://github.com/crc-org/ci-definitions/blob/crc-builder-v1.1.2/crc-builder/tkn/crc-builder-arm64.yaml#L134C1-L140C22

      resources:      
        requests:
          memory: "200Mi"
          cpu: "100m"
        limits:
          memory: "1000Mi"
          cpu: "400m"

but if this is related with those errors (found too on version 0.14.1) worth to review.

Any way to inspect how memory is consumed? Any stable version to test (instead of 0.14.1?)

[adrianriobo]

This task spec does not belong to here, it uses mapt but it is hosted in ci-defintions so you can increase as you wish...can you check an close this issue if it is fixed there?

[albfan]

Just looking for some minimal defaults:

Trying:

mapt/tkn/infra-azure-fedora.yaml

Lines 251 to 257 in aaa5c99

	resources:
	requests:
	memory: "1Gi"
	cpu: "100m"
	limits:
	memory: "2Gi"
	cpu: "300m"

[adrianriobo]

I am totally confused here,

Our tekton task for fedora on Azure already set memory to 2G

mapt/tkn/infra-azure-fedora.yaml

Line 256 in aaa5c99

memory: "2Gi"

you are not using our task but your own which before yesterday it had reduced limit you changed it here crc-org/ci-definitions@c156c8c

This is not a mapt issue you are using the image and you are setting the resources limits...

[albfan]

Yes, that change solve the problem.

Copying the resource assignation on our own task fix it, but is it measurable preventively?

Running locally with

/usr/bin/time -v podman ...

it detects no more memory than 55 Mb consumed. Probably other process consume more RAM resources, how to be sure to do a smooth migration?

[adrianriobo]

migration to what?

[albfan]

Yes, description lacks a little bit of context.

This is a migration from version 0.9.6 to avoid similar problem as #573.

Seing resources at taks at 0.9.6

https://github.com/redhat-developer/mapt/blob/v0.9.6/tkn/infra-azure-fedora.yaml#L194-L200

and updates:

0d78bbe

the guide to reuse these tasks is probably to review diffs in task from versions

[adrianriobo]

So the thing is...either you can use directly our tasks..so upgrade is fine cause we control the resources, or if you use the image is really up to you.

That being said...the path should be update to 0.14.1 but you raised the issue on it so wait a bit I want to have this #842 in and we will release 0.14.2