Join GitHub today
GitHub is home to over 31 million developers working together to host and review code, manage projects, and build software together.Sign up
Ignore ENIs returning 404 on metadata request #194
Issue #, if available:
Description of changes:
This fixes an issue where sometimes new ENIs cannot be attached to an instance,
An example failure scenario:
Click to expand annotated log excerpt showing the issue in action
I confirmed, that this fixes the issue with the following commands:
With these changes, even though I get a
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
I don't get this whole discussion. There is an obvious bug / failure in handling of eventual consistency which seems to be fixed by this pull request. Why suggest all kinds of workarounds instead of merging it? I haven't seen any comment about the pull request itself.
That's not really true. You can just set it to e.g. 20, which will work for all instances, because it's above their max.
True, didn't think of this. The problem with wasted IPs remains.
Can you elaborate on these scenarios? With the current behavior, this is already the case. As long as the the detached ENI is still showing up in the metadata, the IPAMD is basically stuck and doing nothing. New pods still get scheduled on such nodes and can't be started until the ENI disappears from the metadata list and IPAMD continues to work.
referenced this pull request
Jan 31, 2019
For the record, I don't see how that can be the case. The ENIs for which the metadata request fails are ignored by IPAMD with the suggested implementation.