[CK] Add render group to AITER and FA dockers#6563
Merged
Conversation
The AITER and FA test dockers (Dockerfile.aiter, Dockerfile.fa) inherit from rocm/pytorch base image. Recent updates to the base image dropped the render group from /etc/group, which makes docker run fail on test agents with: docker: Error response from daemon: Unable to find group render: no matching entries in group file. This breaks all parallel stages because Jenkins resolves --group-add render against the container's /etc/group, not the host's. Add `groupadd -f render` to both dockers (matching Dockerfile.pytorch and the main Dockerfile) so the group always exists in the container, regardless of base image drift.
DDEle
force-pushed
the
yiding/fix-ck-aiter-fa-docker-render-group
branch
from
80842f5 to
39f4755
Compare
groupadd -f render without -g auto-assigns the next free GID (1001 on the new rocm/pytorch base where only ubuntu=1000 exists), which then collides with the explicit `groupadd -g 1001 jenkins` on the next line: groupadd: GID '1001' already exists Move the render groupadd to after the jenkins user/group is created so render gets 1002 instead. The actual GID render lands on doesn't matter for GPU access because the docker run command also passes `--group-add=109` (the host's render GID) directly.
asleepzzz
approved these changes
Apr 20, 2026
poyenc
approved these changes
Apr 20, 2026
Belt-and-suspenders: current rocm/pytorch base still ships video group (GID 44), but add `groupadd -f video` next to the render one so a future base drop wouldn't break docker run --group-add video either. `-f` is a no-op when the group already exists.
amd-anclark
approved these changes
Apr 20, 2026
assistant-librarian Bot
pushed a commit
to ROCm/composable_kernel
that referenced
this pull request
Apr 21, 2026
[CK] Add render group to AITER and FA dockers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ## Motivation The AITER and FA test dockers (`Dockerfile.aiter`, `Dockerfile.fa`) inherit from the `rocm/pytorch` base image. Recent updates to that base image dropped the `render` group from `/etc/group`, so every parallel test stage now fails on the test agents with: ``` docker: Error response from daemon: Unable to find group render: no matching entries in group file. ``` Jenkins resolves `--group-add render` against the **container's** `/etc/group`, not the host's, so even though the test agents have render in their `/etc/group` (GID 109), the container lookup fails. This pattern affects every recent develop build ([#673](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/673), [#674](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/674), [#686](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/686), [#688](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/688), [#699](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/699), [#708](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/708) — 6 days in a row), where AITER tests fail in seconds and the cascading failure aborts all downstream Build/FMHA/TILE_ENGINE stages. ## Technical Details Add `groupadd -f render` to both `Dockerfile.aiter` and `Dockerfile.fa`, mirroring what the main `Dockerfile` already does (`Dockerfile:96`) and what `Dockerfile.pytorch` does (`Dockerfile.pytorch:4`). The `-f` flag makes it idempotent — silently succeeds if the group already exists. This guarantees the `render` group is always present in the container, regardless of whether the base image happens to ship it. ## Test Plan Triggering AITER CI job: ## Test Result ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
aledudek
pushed a commit
that referenced
this pull request
May 20, 2026
## Motivation The AITER and FA test dockers (`Dockerfile.aiter`, `Dockerfile.fa`) inherit from the `rocm/pytorch` base image. Recent updates to that base image dropped the `render` group from `/etc/group`, so every parallel test stage now fails on the test agents with: ``` docker: Error response from daemon: Unable to find group render: no matching entries in group file. ``` Jenkins resolves `--group-add render` against the **container's** `/etc/group`, not the host's, so even though the test agents have render in their `/etc/group` (GID 109), the container lookup fails. This pattern affects every recent develop build ([#673](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/673), [#674](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/674), [#686](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/686), [#688](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/688), [#699](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/699), [#708](http://micimaster.amd.com/blue/organizations/jenkins/rocm-libraries-folder%2FComposable%20Kernel/detail/develop/708) — 6 days in a row), where AITER tests fail in seconds and the cascading failure aborts all downstream Build/FMHA/TILE_ENGINE stages. ## Technical Details Add `groupadd -f render` to both `Dockerfile.aiter` and `Dockerfile.fa`, mirroring what the main `Dockerfile` already does (`Dockerfile:96`) and what `Dockerfile.pytorch` does (`Dockerfile.pytorch:4`). The `-f` flag makes it idempotent — silently succeeds if the group already exists. This guarantees the `render` group is always present in the container, regardless of whether the base image happens to ship it. ## Test Plan Triggering AITER CI job: ## Test Result ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
The AITER and FA test dockers (
Dockerfile.aiter,Dockerfile.fa) inherit from therocm/pytorchbase image. Recent updates to that base image dropped therendergroup from/etc/group, so every parallel test stage now fails on the test agents with:Jenkins resolves
--group-add renderagainst the container's/etc/group, not the host's, so even though the test agents have render in their/etc/group(GID 109), the container lookup fails.This pattern affects every recent develop build (#673, #674, #686, #688, #699, #708 — 6 days in a row), where AITER tests fail in seconds and the cascading failure aborts all downstream Build/FMHA/TILE_ENGINE stages.
Technical Details
Add
groupadd -f renderto bothDockerfile.aiterandDockerfile.fa, mirroring what the mainDockerfilealready does (Dockerfile:96) and whatDockerfile.pytorchdoes (Dockerfile.pytorch:4). The-fflag makes it idempotent — silently succeeds if the group already exists.This guarantees the
rendergroup is always present in the container, regardless of whether the base image happens to ship it.Test Plan
Triggering AITER CI job:
Test Result
Submission Checklist
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.