github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-05 23:53:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #7260] Migrate off centos 7 for intermediate build layers in container image builds #4614

New Issue
Closed
opened 2026-04-12 15:31:52 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-12 15:31:52 -05:00
Owner
Copy Link

Originally created by @cazlo on GitHub (Oct 18, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/7260

What

Centos is dead, long live centos stream (9)

Ollama should probably not be using centos 7 now that it is unsupported and at EOL.

Why

AMD and Nvidia are no longer publishing updates to their centos 7 flavor of dependencies.

See also https://rocm.docs.amd.com/en/docs-6.2.0/about/release-notes.html

ROCm 6.2.0 marks the end of support (EoS) for:
...
CentOS 7.9

See also https://docs.nvidia.com/cuda/cuda-installation-guide-linux/ not listing centos anywhere.

See also last image Nvidia published for centos 7 was ~6 months ago: https://hub.docker.com/r/nvidia/cuda/tags?name=centos

More info

Currently there are several intermediate build layers in the container image build which utilize centos 7:

Looking at the various Nvidia and AMD docs, it seems like both support the latest EL 9 version, so I would probably try to migrate to EL9 (rockylinux 9 ) to get the latest compatible versions of core dependencies like gcc and also avoid needing to update for a long time (EL 9 EOL is several years away still).

As a quick POC, I was able to get the rocm build migrated to rocky8 with very low effort. This build performance tested the same as the current HEAD of ollama, though I did not run it through the full suite of unit tests.

Originally created by @cazlo on GitHub (Oct 18, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/7260 # What [Centos is dead](https://endoflife.date/centos), long live [centos stream (9)](https://endoflife.date/centos-stream) Ollama should probably not be using centos 7 now that it is unsupported and at EOL. # Why AMD and Nvidia are no longer publishing updates to their centos 7 flavor of dependencies. See also https://rocm.docs.amd.com/en/docs-6.2.0/about/release-notes.html > ROCm 6.2.0 marks the end of support (EoS) for: > ... > CentOS 7.9 See also https://docs.nvidia.com/cuda/cuda-installation-guide-linux/ not listing centos anywhere. See also last image Nvidia published for centos 7 was ~6 months ago: https://hub.docker.com/r/nvidia/cuda/tags?name=centos # More info Currently there are several intermediate build layers in the container image build which utilize centos 7: - [cuda-11-build-amd64](https://github.com/ollama/ollama/blob/bf4018b9ecd56a5deff0c22ca2fba242a8f0101b/Dockerfile#L15) - [cuda-12-build-amd64](https://github.com/ollama/ollama/blob/bf4018b9ecd56a5deff0c22ca2fba242a8f0101b/Dockerfile#L32) - [rocm-build-amd64](https://github.com/ollama/ollama/blob/bf4018b9ecd56a5deff0c22ca2fba242a8f0101b/Dockerfile#L85) - [cpu-builder-amd64](https://github.com/ollama/ollama/blob/bf4018b9ecd56a5deff0c22ca2fba242a8f0101b/Dockerfile#L101) - this one also has transitive layers which depend on it, `container-build-amd64`, `cpu-build-amd64`, `cpu_avx-build-amd64`, and `cpu_avx2-build-amd64`) Looking at the various Nvidia and AMD docs, it seems like both support the latest EL 9 version, so I would probably try to migrate to EL9 (rockylinux 9 ) to get the latest compatible versions of core dependencies like gcc and also avoid needing to update for a long time (EL 9 EOL is several years away still). As a quick POC, I was able to get the rocm build migrated to rocky8 with very low effort. This build performance tested the same as the current HEAD of ollama, though I did not run it through the full suite of unit tests.
GiteaMirror added the buildfeature request labels 2026-04-12 15:31:53 -05:00
GiteaMirror commented 2026-04-12 15:31:54 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Oct 22, 2024):

We've been hesitant to move off Centos 7 for compatibility reasons. For our official linux binaries, the base image we use in the Dockerfile drives the glibc version the binary is linked against, so centos 7 has the broadest compatibility with glibc v2.17. Eventually we do anticipate we'll need to move to a newer base OS, but in doing so, the binary will no longer run on older systems.

<!-- gh-comment-id:2430163079 --> @dhiltgen commented on GitHub (Oct 22, 2024): We've been hesitant to move off Centos 7 for compatibility reasons. For our official linux binaries, the base image we use in the Dockerfile drives the glibc version the binary is linked against, so centos 7 has the broadest compatibility with glibc v2.17. Eventually we do anticipate we'll need to move to a newer base OS, but in doing so, the binary will no longer run on older systems.
GiteaMirror commented 2026-04-12 15:31:54 -05:00
Author
Owner
Copy Link

@cazlo commented on GitHub (Oct 22, 2024):

For what it's worth, for older systems the option to use ollama through the container image exists and should provide the compatibility desired since the container image comes packaged with a known compatible glibc version (and any other dlls).

<!-- gh-comment-id:2430213181 --> @cazlo commented on GitHub (Oct 22, 2024): For what it's worth, for older systems the option to use ollama through the container image exists and should provide the compatibility desired since the container image comes packaged with a known compatible glibc version (and any other dlls).
GiteaMirror commented 2026-04-12 15:31:55 -05:00
Author
Owner
Copy Link

@imweijh commented on GitHub (Feb 12, 2025):

For maximum binary compatibility, we can refer to OpenJDK and use a new operating system along with the sysroot of an older version system:
https://wiki.openjdk.org/display/Build/Supported+Build+Platforms

<!-- gh-comment-id:2653891105 --> @imweijh commented on GitHub (Feb 12, 2025): For maximum binary compatibility, we can refer to OpenJDK and use a new operating system along with the sysroot of an older version system:[ https://wiki.openjdk.org/display/Build/Supported+Build+Platforms]( https://wiki.openjdk.org/display/Build/Supported+Build+Platforms)
GiteaMirror commented 2026-04-12 15:31:55 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 9, 2026):

#8211

<!-- gh-comment-id:3872567211 --> @rick-github commented on GitHub (Feb 9, 2026): #8211
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label build feature request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#4614
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?