-
Notifications
You must be signed in to change notification settings - Fork 615
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
Revendor etcd/raft #1353
Revendor etcd/raft #1353
Conversation
3749751
to
8c1075e
Compare
Current coverage is 55.35% (diff: 100%)@@ master #1353 diff @@
==========================================
Files 80 80
Lines 12637 12632 -5
Methods 0 0
Messages 0 0
Branches 0 0
==========================================
+ Hits 6977 6993 +16
+ Misses 4699 4679 -20
+ Partials 961 960 -1
|
while running this branch, got this:
|
Was this on Mac? I also saw some issues with the tests due to I/O load, but I have an upstream PR pending that should solve the problem. |
Negative. Ubuntu 12.04 LTS. |
2337dd9
to
b03f377
Compare
@dperny: I made a change to the test that may help. Can you check if the problem is still occurring? |
b03f377
to
bce87b0
Compare
My PR to coreos/etcd that allows tests to work around the preallocation issue was merged. Updated this PR and removed WIP. |
@aaronlehmann, sorry, I should have specified that the failure was intermittent. I got it in one run out of maybe 10. |
-22k LOC? I'm in. |
Yeah, some of it is from removing dependencies like candiedyaml that we don't use any more. But also etcd switched to a better vendoring system where they aren't rewriting import paths. So we no longer vendor two copies of gogoprotobuf, for example. |
Change Campaign logic to trigger after Advance, because now Campaign won't do anything if there are uncommitted entries. Also, remove the check on Progress, because now Progress is only filled in if the node is already the leader. Set a smaller WAL segment size to work around high I/O load on Mac when running unit tests. Signed-off-by: Aaron Lehmann <aaron.lehmann@docker.com>
bce87b0
to
62e73cd
Compare
Rebased after #1327 to adapt to new locking API in this upstream version, and adjust PTAL |
Thanks @aaronlehmann, I'm going to test this with docker integration. |
Haven't seen any issue running a cluster with docker integration, removing the progress check seems safe too since we now vendor v3. On 3 Manager cluster, I tested:
We should think about using the leadership transfer on self-demotion now that we vendor v3. LGTM |
Cool. Did you have a chance to run the Docker integration tests? I haven't done this yet. |
Docker integration tests passed. Here goes nothing. |
@LK4D4: Actually it's -36k LOC - GitHub is not showing correct totals. |
Change Campaign logic to trigger after Advance, because now Campaign
won't do anything if there are uncommitted entries. Also, remove the
check on Progress, because now Progress is only filled in if the node is
already the leader.
Set a smaller WAL segment size to work around high I/O load on Mac when
running unit tests.
Fixes #1229