Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CentOS fluent-bit packages repository is not indexing new packages #7016

Closed
braunsonm opened this issue Mar 14, 2023 · 7 comments · Fixed by fluent/fluent-bit-infra#183
Closed
Assignees
Labels

Comments

@braunsonm
Copy link

Bug Report

Describe the bug
Fluent-bit rpm's are released in the CentOS repository, but it looks like the repodata (which dnf, yum use) has not been updated correctly so the operating system is unaware there is an update.

To Reproduce

  • Go to https://packages.fluentbit.io/centos/7/x86_64/repodata/
  • If you download the repomd.xml you will see the primary database that is being referenced. This is an old index from January and is not the newest which includes 2.0.9. It looks like the new index should be https://packages.fluentbit.io/centos/7/x86_64/repodata/e7f0d5f9ce01be8238af0127051bf0b760cdadc18155e210bfe5ce2e0fb4089c-primary.xml.gz but it is not being referenced right now.

Expected behavior
The CentOS package repository is correctly updated.

@patrick-stephens
Copy link
Contributor

I think this is an artefact of the transition from the old release server which had a different set up that unfortunately had to be replicated for the docs so has some delightful symlinks.

Testing with a local install (please don't rely on what the browser shows) shows it pulling 2.0.8 instead of 2.0.9 so is indeed wrong, I will take a look.

@patrick-stephens
Copy link
Contributor

I will check the server has sync'd from the S3 bucket correctly. We did have an issue with Cloudflare caching the metadata but that should be resolved now.

@patrick-stephens
Copy link
Contributor

patrick-stephens commented Mar 15, 2023

There was a failure due to disk space on the runner for the 2.0.9 release which meant I had to retry the repo update so wondering if that is related?

Separately we have a test to verify we can install the correct version of the package after release but this works with the S3 bucket directly so makes me suspect an issue with the sync to the server: https://github.com/fluent/fluent-bit/actions/runs/4102328393/jobs/7076569800

@patrick-stephens
Copy link
Contributor

Yeah, the S3 bucket seems to have it correct:

$ curl http://packages.fluentbit.io.s3.amazonaws.com/centos/7/repodata/repomd.xml
<?xml version="1.0" encoding="UTF-8"?>
<repomd xmlns="http://linux.duke.edu/metadata/repo" xmlns:rpm="http://linux.duke.edu/metadata/rpm">
  <revision>1675680680</revision>
  <data type="primary">
    <checksum type="sha256">e7f0d5f9ce01be8238af0127051bf0b760cdadc18155e210bfe5ce2e0fb4089c</checksum>
    <open-checksum type="sha256">fda8fadda86b31086c00d7c4505f64cc4baddd7facf91bf517c02a25f14762bc</open-checksum>
    <location href="repodata/e7f0d5f9ce01be8238af0127051bf0b760cdadc18155e210bfe5ce2e0fb4089c-primary.xml.gz"/>
    <timestamp>1675680680</timestamp>
    <size>38971</size>
    <open-size>844442</open-size>
  </data>

vs the server sync'd version:

$ curl https://packages.fluentbit.io/centos/7/repodata/repomd.xml
<?xml version="1.0" encoding="UTF-8"?>
<repomd xmlns="http://linux.duke.edu/metadata/repo" xmlns:rpm="http://linux.duke.edu/metadata/rpm">
  <revision>1673886886</revision>
  <data type="primary">
    <checksum type="sha256">577f465a88688aecb0d1d2e72537ae9a687b15d533ab3565b056da5f25296513</checksum>
    <open-checksum type="sha256">390b1c86eb8181f506ad35b492db6f7dd1281c43fbc2ba144eae91bcb6d2ece3</open-checksum>
    <location href="repodata/577f465a88688aecb0d1d2e72537ae9a687b15d533ab3565b056da5f25296513-primary.xml.gz"/>
    <timestamp>1673886886</timestamp>
    <size>38425</size>
    <open-size>833407</open-size>
  </data>

So, looks like the release built correctly and did everything but the aws s3 sync has not updated the contents on the server side. I may have to force it to for metadata depending on how the aws client decides a file has changed. I'll also check that Cloudflare is not caching it and providing a cached copy.

@patrick-stephens
Copy link
Contributor

This was a fun ride, turns out the aws sync ignores timestamps locally unless you force it: aws/aws-cli#1074

Metadata files do not change size generally so required me to add the --exact-timestamps flag to get it working on the packages server.

@patrick-stephens
Copy link
Contributor

@braunsonm please confirm but I was reproducing it before and now it is fine.

@braunsonm
Copy link
Author

Can confirm it is indexed properly now @patrick-stephens

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging a pull request may close this issue.

2 participants