dotnet
Remove CentOS and RedHat from the Helix environment
CentOS and RedHat meet the EOL date on 2024-06-30, once the customers stop using them, they should be removed.
Release Note Category
- [x] Feature changes/additions
- [ ] Bug fixes
- [ ] Internal Infrastructure Improvements
Release Note Description
Removed the queues for CentOS and RedHat i.e., all centos.7.amd64* and redhat.7.amd64* queues.
We need to work with the product teams to make sure workloads are moved to other queues before these are removed.
CentOS 7 is in a worse state than RedHat 7 b/c RedHat at least has an extended support option for RedHat 7. CentOS 7 is completely EOL as of 30 June '24. I'm pretty sure we can't create CentOS 7 images anymore. Will test…
A few details
Performance, runtime, and wcf runtimes still use the centos7.amd64[.open], even over the past week.
See Teams conversation and Delete CentOS 7 Dockerfiles PR reasoning. Note workarounds for missing package repo mirror choice site don't make new package updates available for the OS.
I'm pretty sure we can't create CentOS 7 images anymore. Will test…
Turns out image generation has no problems installing CentOS 7 packages though one of the RedHat changes to "incentivize" moving off that OS was to disable the service used to choose between package repository mirrors. See !40896, success of MakeImage operations in #20240706.01, and agent logs in Image Factory run 9b16eda7-4b1b-5f13-94c9-0592dab9de75.
In addition, execute in [Web] [Desktop] [cluster('engsrvprod.kusto.windows.net').database('engineeringdata')]
Jobs
| where QueueAlias has "centos" and Finished > ago(5d)
| project JobId, Source, Type, Queued, Started, Finished, QueueName, Repository, Branch
shows numerous repositories and their pipelines continue to test on CentOS queues.
Above Kusto query mixed up QueueAlias and QueueName. Correct version is
Execute in [Web] [Desktop] [cluster('engsrvprod.kusto.windows.net').database('engineeringdata')]
Jobs | where QueueName has "centos" and Finished > ago(30d) | project JobId, Source, Type, QueueName, Repository, Branch, Queued, Started, Finished
or, for redhat: Execute in [Web] [Desktop] [cluster('engsrvprod.kusto.windows.net').database('engineeringdata')]
Jobs | where QueueName has "RedHat" and Finished > ago(30d) | project JobId, Source, Type, QueueName, Repository, Branch, Queued, Started, Finished
Current data shows performance team has cleaned up their remaining use of the CentOS queue but RedHat use continues in at least aspnetcore (for release/6.0), arcade, and runtime pipelines.
Not sure how to resolve this w/o another discussion w/ .NET PMs or waiting for @ilyas1974 to return. One idea might be to remove the CentOS queue and start a conversation about RedHat. @garath and @markwilkie what are your thoughts❓