-
Notifications
You must be signed in to change notification settings - Fork 1.5k
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
Default S3 Setup Documentation Results in Broken Config for EKS #4654
Comments
I just had my cluster scale and I have not updated the launch template to set the hop count, but I did add the AWS filter with IMDS v1 as the setting. So for the logs below: IMDS tokens are set to Optional
Unfortunately the containers on the new instance still hang on startup, but they have a different log output pattern:
|
One other comment. I used the exact same S3 OUTPUT config, with no AWS filter, in the same cluster running Fluent Bit 1.8.1 (via New Relic install) and it works completely fine. |
Also seeing this in EKS. I've downgraded fluentbit significantly to avoid this. |
This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 5 days. Maintainers can add the |
This issue was closed because it has been stalled for 5 days with no activity. |
Bug Report
Describe the bug
Using the latest FluentBit Helm Chart I deployed FluentBit 1.8.11 to an EKS cluster and it ran great. I then added the S3 output and noticed that the container was in a CrashLoopBackoff state.
To Reproduce
Expected behavior
The HTTP server should start successfully - OR - an error of some kind should be output in the logs.
Screenshots
Your Environment
Additional context
I was trying to push logs to S3 for archiving, but the solution wouldn't work.
Caveat: This very likely may be a documentation issue, but it does seem like some code updates would make this more user friendly.
I found, after reading several GitHub issues, a curl command for using IMDSv2. I tried the command inside an ElasticSearch pod in the same cluster and that command hung (the default fluent-bit containers don't have /bin/sh nor /bin/bash). I have not previously made IMDS required on our instances but rather made it Optional.
I then found commentary stating that a fix for IMDSv2 in Fluent-bit 1.8.8 was needed by changing the hop count. I am on 1.8.11 so didn't think this was required (somewhere I saw a comment saying the issue was fixed in 1.8.9), but I changed the setting anyway. Suddenly the pods started working. I then learned about the AWS Filter from a GitHub issue as well and read the docs there.
I noticed that the default setting for AWS is to use IMDSv2 which seems to require a token, but if I have IMDSv2 as optional, the aforementioned curl command when the hop count is 1 just HANGS.
So either a timeout on the IMDSv2 (and add a corresponding
initialDelaySeconds
to both thelivenessProbe
and thereadinessProbe
to the Helm chart) with an error message and/or update the S3 documentation to state caveats about the IMDS version settings and link to the AWS filter documentation.If I can help some how, let me know.
The text was updated successfully, but these errors were encountered: