Add host-local IPAM GC on startup #5660
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
/test-all |
antoninbas
force-pushed
the
cniserver-ipam-gc-2
branch
from
29e53b5 to
3b3c873
Compare
antoninbas
force-pushed
the
cniserver-ipam-gc-2
branch
from
3b3c873 to
4bb14d3
Compare
antoninbas
added
action/release-note
tnqn
reviewed
Nov 14, 2023
| return fmt.Errorf("path '%s' is not a directory: %w", dir, err) | ||
| } | ||
|
|
||
| lk, err := NewFileLock(dataDir) |
There was a problem hiding this comment.
It seems https://github.com/containernetworking/plugins/blob/93a1b3d0e71b3a4b5c463de357876fe8d383012f/plugins/ipam/host-local/backend/disk/backend.go#L53 uses dir instead of its parent dir dataDir?
There was a problem hiding this comment.
That's a pretty bad typo, fixed
| return fmt.Errorf("error when gathering IP filenames in the host-local data directory: %w", err) | ||
| } | ||
|
|
||
| allocatedIPs := sets.New[string]() |
There was a problem hiding this comment.
nit: it seems enough to use a bool to track if it ever fails to release an IP.
| "github.com/alexflint/go-filemutex" | ||
| ) | ||
|
|
||
| // This code was copied from https://github.com/containernetworking/plugins/blob/v1.3.0/plugins/ipam/host-local/backend/disk/lock.go |
There was a problem hiding this comment.
could we import github.com/containernetworking/plugins/plugins/ipam/host-local/backend given the package is already a dependency?
Line 22 in 4a52bca
There was a problem hiding this comment.
Yes, we actually use it pretty extensively. Changed to an import.
During CNIServer reconciliation, we perform host-local IPAM garbage collection (GC) by comparing the set of IPs allocated to local Pods and the set of IPs currently reserved by the plugin. We release any IP reserved by the plugin that is not in-use by a local Pod. The purpose is to avoid leaking IP addresses when there is a bug in the container runtime, which has happened in the past. Two key design choices that were made: * We do not invoke CNI DEL to release IPs, instead we access the host-local data which is persisted on the Node, and modify it as needed. * We do not rely on the interface store (as persisted to OVSDB) to determine the set of IPs that may have been leaked. In case of an Antrea bug, it could be possible (although unlikely) for an IP to still be allocated by host-local but be missing from the interface store. Intead, we list all allocated IPs from the host-local data (an allocated IP corresponds to one disk file). This approach is essentially the same as our existing script: https://github.com/antrea-io/antrea/blob/main/hack/gc-host-local.sh Fixes antrea-io#4326 Signed-off-by: Antonin Bas <abas@vmware.com>
antoninbas
force-pushed
the
cniserver-ipam-gc-2
branch
2 times, most recently
from
37a66c4 to
23c2725
Compare
antoninbas
force-pushed
the
cniserver-ipam-gc-2
branch
2 times, most recently
from
6b33047 to
17f523c
Compare
Signed-off-by: Antonin Bas <abas@vmware.com>
antoninbas
force-pushed
the
cniserver-ipam-gc-2
branch
from
17f523c to
4223995
Compare
|
/test-all |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
During CNIServer reconciliation, we perform host-local IPAM garbage collection (GC) by comparing the set of IPs allocated to local Pods and the set of IPs currently reserved by the plugin. We release any IP reserved by the plugin that is not in-use by a local Pod. The purpose is to avoid leaking IP addresses when there is a bug in the container runtime, which has happened in the past.
Two key design choices that were made:
This approach is essentially the same as our existing script: https://github.com/antrea-io/antrea/blob/main/hack/gc-host-local.sh
Fixes #4326