Deal with auto-egress-ip mark conflicting with kube-proxy's masqueradeBit #18121
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
danwinship
added
kind/bug
openshift-ci-robot
added
the
size/L
|
/retest |
|
/lgtm |
|
@dcbw PTAL |
openshift-merge-robot
added
the
needs-rebase
danwinship
force-pushed
the
egress-ip-mark-conflicts
branch
from
721dd58 to
c682166
Compare
openshift-merge-robot
removed
lgtm
danwinship
force-pushed
the
egress-ip-mark-conflicts
branch
from
c682166 to
cceabf8
Compare
|
/lgtm |
|
/approve |
|
/assign @smarterclayton Clayton, PTAL... we need sign-off for cmd again. Thanks! |
|
/lgtm |
|
@smarterclayton can you approve us for the /pkg/cmd change please? |
|
/approve |
openshift-ci-robot
added
the
approved
|
/retest Please review the full test history for this PR and help us cut down flakes. |
1 similar comment
|
/retest Please review the full test history for this PR and help us cut down flakes. |
openshift-merge-robot
added
the
needs-rebase
If the iptables rules get reordered after a flush, then it's possible that kube-proxy rules that look at MasqueradeBit will accidentally match a packet that was intended for an auto-egress-ip rule. To avoid that, change the code to avoid using that bit of pkt_mark. This means we now only have 31 bits to work with instead of 32, so make the mark be based on the (24-bit) VNID rather than the (32-bit) egress IP address.
danwinship
force-pushed
the
egress-ip-mark-conflicts
branch
from
cceabf8 to
d8ec67e
Compare
openshift-merge-robot
removed
lgtm
|
/retest |
|
/lgtm |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: danwinship, dcbw, knobunc, smarterclayton The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these OWNERS Files:
You can indicate your approval by writing |
|
Automatic merge from submit-queue. |
openshift-merge-robot
added a commit
that referenced
this pull request
Feb 2, 2018
Automatic merge from submit-queue. Drop auto-egress-IP rules when egress IP is removed from NetNamespace (Previously we were only doing it when the NetNamespace was deleted.) Fixes https://bugzilla.redhat.com/show_bug.cgi?id=1540846 (does not conflict with #18121 so it doesn't matter which merges first)
openshift-merge-robot
added a commit
that referenced
this pull request
Feb 23, 2018
Automatic merge from submit-queue. Fix reassignment of egress IP after removal When dropping an egress IP from eth0, we weren't updating our internal state to reflect that we had done that, so if you added it back again it wouldn't do everything it needed to do. (Introduced in #18121 but not discoverable until after #18547 was fixed.) Fixes https://bugzilla.redhat.com/show_bug.cgi?id=1547899
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.
auto-egress-ip uses the skb mark (OVS
pkt_mark, iptablesmark) to communicate information from OVS rules to iptables rules. Normally this works fine, but if the iptables rules get flushed and recreated, it's possible that the OpenShift and kube-proxy rules will get added back in the "wrong" order, and then the kube-proxy rules will start matching some of the auto-egress-ip packets, causing the wrong thing to happen.Forcibly fixing the iptables rule order is difficult and fragile. Changing the implementation to not use the mark would also not be easy (and would be a lot of code to backport to 3.7). So I fixed it by just changing the code to use mark values that don't conflict with the bit that kube-proxy is using.
This means we only have 31 bits of mark to work with, so we can't reliably use the IP address as the mark value any more. So I switched to using the namespace's VNID instead, since those are only 24 bits. But this required reorganizing the code a bit because previously we were setting up the iptables rules for egress IPs before we knew what namespaces they were associated with.
First commit ("Make sure oc.tunMAC gets set even if AlreadySetUp()") is from #18049
Fixes https://bugzilla.redhat.com/show_bug.cgi?id=1527642
@openshift/sig-networking PTAL