-
Notifications
You must be signed in to change notification settings - Fork 3.4k
ipam/aws: update the assigned IPs from aws to instance map #37650
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
ipam/aws: update the assigned IPs from aws to instance map #37650
Conversation
/test |
bd54d82
to
054974f
Compare
/test |
054974f
to
ad5066d
Compare
/test |
ad5066d
to
dabe5cd
Compare
I had the commit to print out the result to validate the result(print out the IP address getting from AWS and print out the IP in the ENI from before and after using
I guess we should do similiar thing for prefix delegation but I will raise another PR for it later once we agree this is the best way to resolve the issue |
/test |
/ci-ipsec-e2e |
/ci-ipsec-upgrade |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't see anything wrong in this PR but since @gandro was involved in the discussion, it would be nice to have his review.
Apologies for the delay in review. Since (at least to my knowledge) we still don't understand the root cause of the stale data in the instances manager, I'm a bit worried that this papers over a real issue that might affect other API calls (e.g. attaching a new ENI) as well. Edit: Discussed offline, apparently attaching a new ENI already uses this approach: Lines 535 to 536 in e21d1e4
Having said that, if we do not find the real root cause, I'm happy to merge this as a workaround. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I did some more thinking about this approach, and I now agree that it is overall the right thing to do.
There is one potential race that I think needs to be fixed, but otherwise I think this will improve the situation of us miscalculating the number of available addresses.
dabe5cd
to
4bd1599
Compare
/test |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks!
Note that this needs a rebase for CI to be able to pass https://cilium.slack.com/archives/C7PE7V806/p1740452895682009 |
Great. I'm waiting for this PR get merged so I can work on my orginal PR(#36922). I will also have the follow up PR to remove the IP and dealing with PD. Shall I open other issues? or I should open the agianst #36428 |
4bd1599
to
3411e23
Compare
/test |
/ci-clustermesh |
/ci-integration |
/ci-awscni |
3411e23
to
e4b6f0b
Compare
/test |
I re-request a review since I have a new commit but I don't have the permission to remove the ready to merge label. Please don't merge until I get a new approval |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks! Remove logic looks good to me! One subjective non-blocking nit
- Extracting and returning the assigned IP addresses - Adding new methods to update ENI IP addresses in the instances manager - Update the IP to ENI after getting the IP from aws Signed-off-by: Liyi Huang <liyi.huang@isovalent.com>
e4b6f0b
to
0801cff
Compare
/test |
this is the temporary fix for ENI mode to avoid the sync from AWS getting the stale data since we would like to see if this can fix the CI issue for cilium#36428 based on the discussion here cilium#36428 (comment) if so, we will change how alibaba and azure behavior like PR cilium#37650 and remove this instanceSync trigger sync from here completely. I have tested skip this trigger sync with draft PR cilium#40760 for a few times and it seems to be working. we hope we can test this for 2-4 weeks so we can know if the CI issue got fixed. if we dont see the CI issue again, we will have the follow up PR to change Azure and Alibaba cloud and remove this trigger. Signed-off-by: Liyi Huang <liyi.huang@isovalent.com>
this is the temporary fix for ENI mode to avoid the sync from AWS getting the stale data since we would like to see if this can fix the CI issue for cilium#36428 based on the discussion here cilium#36428 (comment) if so, we will change how alibaba and azure behavior like PR cilium#37650 and remove this instanceSync trigger sync from here completely. I have tested skip this trigger sync with draft PR cilium#40760 for a few times and it seems to be working. we hope we can test this for 2-4 weeks so we can know if the CI issue got fixed. if we dont see the CI issue again, we will have the follow up PR to change Azure and Alibaba cloud and remove this trigger. Signed-off-by: Liyi Huang <liyi.huang@isovalent.com>
this is the temporary fix for ENI mode to avoid the sync from AWS getting the stale data since we would like to see if this can fix the CI issue for cilium#36428 based on the discussion here cilium#36428 (comment) if so, we will change how alibaba and azure behavior like PR cilium#37650 and remove this instanceSync trigger sync from here completely. I have tested skip this trigger sync with draft PR cilium#40760 for a few times and it seems to be working. we hope we can test this for 2-4 weeks so we can know if the CI issue got fixed. if we dont see the CI issue again, we will have the follow up PR to change Azure and Alibaba cloud and remove this trigger. Signed-off-by: Liyi Huang <liyi.huang@isovalent.com>
Fixes: #36428