Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
That seems reasonable to me, I'll provide a separate PR for improving the
connection loss thing in ZOOKEEPER-3157.
---
Github user anmolnar commented on the issue:
https://github.com/apache/zookeeper/pull/605
+1 for @hanm 's concern.
Auto-retrying tests wouldn't be good here.
---
Github user hanm commented on the issue:
https://github.com/apache/zookeeper/pull/605
>> since it's actually 'hiding' the flaky tests
This is my main concern. If we do this to all test cases, it's very likely
there will be no follow ups or investigations on the flaky tests
Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
@anmolnar thanks for the context, I'll provide a separate patch for 3.5 as
well.
Also I created a JIRA to improve the test: ZOOKEEPER-3157.
As I mentioned in that JIRA,
Github user anmolnar commented on the issue:
https://github.com/apache/zookeeper/pull/605
Ooops, sorry @hanm , you're a little bit late, this one is already
committed.
Would you or @lvfangmin mind creating a separate Jira for the flaky test
fix?
---
Github user anmolnar commented on the issue:
https://github.com/apache/zookeeper/pull/605
@lvfangmin Errm. I would be happy to do that, but not sure how to. The
merge script is able to cherrry-pick in the middle of the process, but I cannot
start it over without a pull request.
Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
@anmolnar after #631 being merged into 3.5, this patch can be cherry-picked
back to 3.5 without conflict, can you help cherry-pick it back?
---
Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
@anmolnar I created a PR to port ZOOKEEPER-3127 to 3.5: #631, I'll port
ZOOKEEPER-3125 back after that, because 3125 is depending on 3127.
---
Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
@anmolnar, sorry, just saw your comment, I'll send out a PR for 3.5 today.
---
Github user anmolnar commented on the issue:
https://github.com/apache/zookeeper/pull/605
@lvfangmin Cool, thanks for the info, but unfortunately the commit script
is quite strict and works with PRs. Maybe I could circumvent the process and
commit the change manually, but I think it
Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
@anmolnar the conflict is due to the missing of FuzzySnapshotRelatedTest
class introduced in ZOOKEEPER-3127, which is fixing the data inconsistency due
to multi-op txn.
We need that in
Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
Thanks @anmolnar, will create a PR for 3.5.
---
Github user anmolnar commented on the issue:
https://github.com/apache/zookeeper/pull/605
@lvfangmin Committed to master. Conflicting with branch-3.5, please create
separate pull request for that.
---
Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
Just reported another pzxid inconsistent issue related to CloseSession, I
found that issue recently, will wait this being merged before providing that
patch, which will make adding the unit
Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
Just found there is conflict to merge, I'm updating it now.
Btw, I just found 2 other potential data inconsistent issues about multi-op
and ephemeral, will create Jiras and send out
Github user lvfangmin commented on the issue:
https://github.com/apache/zookeeper/pull/605
This could cause watch event not notified when setWatches because of stale
pzxid, and all the tests have been passed, I think we should land this and port
it back to other branches as early as
16 matches
Mail list logo
