OCTRL-1078 Make Task Controller able to control OCC tasks by justonedev1 · Pull Request #804 · AliceO2Group/Control

justonedev1 · 2026-03-17T16:40:40Z

introduction of controller for kubernetes inside the controller-operator folder.

This version allows us to control all OCC tasks (tested on readout, stfsender and stfbuilder). Currently there is not image and controller was run directly via make run on the node where the kubernetes cluster is running. If you run it locally and cluster is on different computer, you would not be able to communicate with OCC process as these require opened ports
Manifests required to run these tasks are inside the ecs-manifests subfolder.

In order to apply these manifests use kubectl. There is an explanation which file is which inside the kubernetes-ecs.md.

… OCC tasks

knopers8

thanks! I did not try to run it yet, but I already have some questions/suggestions.

knopers8 · 2026-03-19T13:29:36Z

control-operator/api/v1alpha1/groupversion_info.go

+ * === This file is part of ALICE O² ===
+ *
+ * Copyright 2024 CERN and copyright holders of ALICE O².
+ * Author: Teo Mrnjavac <teo.mrnjavac@cern.ch>


Suggested change

* Author: Teo Mrnjavac <teo.mrnjavac@cern.ch>

applies to the rest as well.

knopers8 · 2026-03-19T13:34:27Z

control-operator/cmd/main.go

+/*
+Copyright 2024.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/


just a thought, we might want to know whether we should keep the original licences of auto-generated files or move to the one we use in O2.

knopers8 · 2026-03-19T13:39:33Z

control-operator/config/manager/kustomization.yaml

+kind: Kustomization
+images:
+- name: controller
+  newName: docker.io/teom/aliecs


that's valid? see also the Makefile which refers to teom

yeah, I haven't tried to build and deploy docker image by then yet, I was running everything by make run. I changed the names. However, in my upcoming PR with environment controller, I changed and split kustomization for each controller and fixed this naming.

knopers8 · 2026-03-19T13:45:22Z

control-operator/ecs-manifests/control-workflows/readout-kube.yaml

+    - O2_PARTITION={{ environment_id }}
+  user: "{{ user }}"
+  arguments:
+    - "/root/readout.yaml"


why is the manifest file added as an argument to readout?

it is described in the documentation, basically it is a scaffold to let the executor know which manifest to load. You can see it inside the KubectlTask in other PR

knopers8 · 2026-03-19T14:14:04Z

control-operator/ecs-manifests/control-workflows/stfbuilder-senderoutput-kube.yaml

+  user: "{{ user }}"
+  value: "{{ len(modulepath)>0 ? _module_cmdline : _plain_cmdline }}"
+  arguments:
+    - "/root/stfbuilder-senderoutput.yaml"


same question here

knopers8 · 2026-03-19T14:43:29Z

control-operator/internal/controller/fairmq_transition.go

+)
+
+// fmqToOCCState maps FairMQ internal states to lowercase OCC state names
+var fmqToOCCState = map[string]string{


FYI, we do it a bit differently in the current system: https://github.com/AliceO2Group/Control/blob/master/core/integration/odc/fairmq/fairmq.go

It probably doesn't matter much in here though.

I don't understand how is propagation of states from ODC to ecs core relevant in a context of sending and receiveing changes from/to fairmq devices

The states from ODC are the states of FairMQ devices.

sorry, I misunderstood what you are objecting to. You are right, I fixed it according to ODC rules, so we can at least see if there is some problem during multi-step transitions. But over-all it shouldn't matter too match as you mentioned in the original comment

knopers8 · 2026-03-19T15:06:23Z

control-operator/exampletask-test.yaml

@@ -0,0 +1,69 @@
+apiVersion: aliecs.alice.cern/v1alpha1


Perhaps this and the other manifest in this directory could be moved to ecs-manifests.

this is leftover

knopers8 · 2026-03-19T15:10:21Z

control-operator/Makefile

@@ -0,0 +1,306 @@
+# VERSION defines the project version for the bundle.


for my curiousity, this file is auto-generated by the operator SDK or AI-generated?

this is all, first version is generated, than I made some changes manually and also used llm

knopers8 · 2026-03-19T15:14:40Z

control-operator/internal/controller/task_controller.go

+	log.V(1).Info("new reconcile request on existing Task Kind", "request", req)
+
+	// Handle finalizers for clean deletion
+	if res, stop, err := r.handleFinalizer(ctx, t, log); err != nil || stop {


i am struggling to understand what it is needed for, could you explain please?

finalizer is used for cleaning up tasks... it is basically just a tag to say "wait don't delete yet"
https://kubernetes.io/docs/concepts/overview/working-with-objects/finalizers/

knopers8 · 2026-03-19T15:30:22Z

control-operator/internal/controller/task_controller.go

+	// As far as I understand this is necessary because even though we have gRPC streams, we cannot rely
+	// on them being implemented


What do you mean that they might not be implemented? Why do we recognize it by seeing an empty t.Status.State?

occlite doesn't implement StateStream, so we cannot rely on the consuming it. and get the status via automatic update, we need to query it.

justonedev1 requested a review from knopers8 as a code owner March 17, 2026 16:40

justonedev1 force-pushed the OCTRL-1078 branch from d7944b9 to b4bcee7 Compare March 18, 2026 12:42

[operator] OCTRL-1078 controller operator that is able to control all…

8421f88

… OCC tasks

justonedev1 force-pushed the OCTRL-1078 branch from b4bcee7 to 8421f88 Compare March 18, 2026 12:47

knopers8 reviewed Mar 19, 2026

View reviewed changes

Michal Tichák added 3 commits April 9, 2026 15:56

[operator] fairmq states are better mapped

b18f26d

[operator] changed name in copyright

aee9cf5

[operator] moved to correct image name

e14c7a6

justonedev1 requested a review from knopers8 April 9, 2026 14:22

		@@ -0,0 +1,306 @@
		# VERSION defines the project version for the bundle.

		// As far as I understand this is necessary because even though we have gRPC streams, we cannot rely
		// on them being implemented

Conversation

justonedev1 commented Mar 17, 2026

Uh oh!

knopers8 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

knopers8 Apr 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

knopers8 Apr 9, 2026 •

edited

Loading