Remove watchMiss for swarm mode

[fcrisciani]

Swarm mode does not really have anymore a use for the watchMiss.
Peer entries are configured at configuration time.
If the gcthresh denies the insertion the peerAdd will fail.

Signed-off-by: Flavio Crisciani flavio.crisciani@docker.com

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@83862f4). Click here to learn what that means.
The diff coverage is 0%.

@@ Coverage Diff @@## master #2047 +/- ##
=========================================
Coverage ? 40.07% =========================================
Files ? 138 Lines ? 22126 Branches ? 0 =========================================
Hits ? 8866 Misses ? 11961 Partials ? 1299

Impacted Files	Coverage Δ
drivers/overlay/ov_network.go	2.18% <0%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 83862f4...3d9a808. Read the comment docs.

[selansen]

LGTM

[ddebroy]

LGTM.

To make sure I understand the context here: we thought we needed this notification to reprogram neighbor ARP entries (#1842) but it was later found that the reprogramming was unnecessary (49200cb#diff-50ec864109e0293aa03f973d35485e1dR778) . So to save on netlink resources, we are getting rid of the watch that was kept around from the initial changes, correct?

[fcrisciani]

@ddebroy the premise is that this optimization relates only to swarm mode, swarm classic behavior is untouched.

In the past instead of fixing the real root cause that was ending up with a wrong configuration functions like this got introduced with the aim to basically remediate re writing the peer entry if not present. The peer operation are done at configuration time not at run time.
With the work that we did in the past months we identified several race conditions that were the cause for a miss alignment into the peer configuration on the overlay namespace so basically at this point the function is merely writing a log if there is a missing entry.
During some tests I noticed how expensive is this function as well as not super useful for its purpose so imo can go :)

[selansen]

This looks like pretty good clean up :)

…

On Mon, Jan 8, 2018 at 7:48 PM, Flavio Crisciani ***@***.***> wrote: @ddebroy <https://github.com/ddebroy> the premise is that this optimization relates only to swarm mode, swarm classic behavior is untouched. In the past instead of fixing the real root cause that was ending up with a wrong configuration functions like this got introduced with the aim to basically remediate re writing the peer entry if not present. The peer operation are done at configuration time not at run time. With the work that we did in the past months we identified several race conditions that were the cause for a miss alignment into the peer configuration on the overlay namespace so basically at this point the function is merely writing a log if there is a missing entry. During some tests I noticed how expensive is this function as well as not super useful for its purpose so imo can go :) — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2047 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKEKn9u6M2J4hOVtugOpuIb6kOiytWhUks5tIrdGgaJpZM4RVFCc> .