Got a new failure:
RATIS-2108 testLeaderLeaseDuringReconfiguration may fail with
IllegalStateException: allLeaks.size = 1
Interestingly, I got "IllegalStateException: allLeaks.size = ..." even if I
did not enable zero copy.
*+++
b/ratis-grpc/src/test/java/org/apache/ratis/grpc/MiniRaftClusterWithGrpc.java*
@@ -77,7 +77,7 @@ public class MiniRaftClusterWithGrpc extends
MiniRaftCluster.RpcBase {
Optional.ofNullable(getAddress(id, group,
RaftPeer::getAdminAddress)).ifPresent(address ->
GrpcConfigKeys.Admin.setPort(properties,
NetUtils.createSocketAddr(address).getPort()));
// Always run grpc integration tests with zero-copy enabled because
the path of nonzero-copy is not risky.
- GrpcConfigKeys.Server.setZeroCopyEnabled(properties, true);
+// GrpcConfigKeys.Server.setZeroCopyEnabled(properties, true);
return parameters;
}
Tsz-Wo
On Tue, Jun 4, 2024 at 9:24 AM Tsz Wo Sze <[email protected]> wrote:
> Hi William and Ashish,
>
> For RATIS-2104, Duong found out that the problem was the test itself,
> which shut down the cluster twice.
>
> RATIS-2105 is for NettyRpc. It seems okay since we have not changed
> NettyRpc.
>
> I will rerun the tests one more time.
>
> Tsz-Wo
>
>
>
> On Mon, Jun 3, 2024 at 7:43 PM Ashish <[email protected]> wrote:
>
>> These are test failures, so we can create JIRA's and monitor them.
>>
>> Are we planning to hold the vote for these failures, since we already
>> passed 72 hr window ?
>>
>> On Mon, Jun 3, 2024 at 10:31 AM Tsz Wo Sze <[email protected]> wrote:
>>
>> > Hi William,
>> >
>> > I am checking the failures in more details to see if they are real
>> > problems.
>> >
>> > Tsz-Wo
>> >
>> > On Mon, Jun 3, 2024 at 4:19 PM William Song <[email protected]> wrote:
>> >
>> > > Hi Tsz-Wo,
>> > >
>> > > Thanks for verifying the release candidate, will keep an eye on
>> > > RATIS-2104, 2105.
>> > >
>> > > I’m trying to reproduce the test failure on my local PC (Mac M chip,
>> JDK
>> > > 8/JDK 20) but haven’t see those failures. For RATIS-2104, during the
>> > > development of TestLeaderInstallSnapshot I've also seen the
>> > > java.lang.IllegalStateException (allLeaks.size=5), suggesting
>> > > AppendEntriesProtos are leaked, if I forget to let the isolated leader
>> > join
>> > > back at the end.
>> > >
>> > > I’ll keep on reproducing RATIS-2104, 2105.
>> > >
>> > > William
>> > >
>> > > > 2024年6月3日 07:57，Tsz Wo Sze <[email protected]> 写道：
>> > > >
>> > > > Similar to Ashish, I also saw some test failures. The following two
>> > > tests
>> > > > failed all the time.
>> > > > - RATIS-2104 TestLeaderInstallSnapshot may fail with
>> > > > java.lang.IllegalStateException: allLeaks.size = 4
>> > > > - RATIS-2105 TestRetryCacheWithNettyRpc may fail with
>> TimeoutException
>> > > >
>> > > > Will check the details.
>> > > >
>> > > > Tsz-Wo
>> > > >
>> > > > On Sat, Jun 1, 2024 at 11:06 PM Tsz Wo Sze <[email protected]>
>> wrote:
>> > > >
>> > > >> I am still running tests. Will vote soon.
>> > > >>
>> > > >> @Ashish, thanks for verifying the release and vote!
>> > > >>
>> > > >>> Side note, on the 1st try I did get test failure due to timeout.
>> > > >>
>> > > >> Occasionally, tests may fail with a bind exception which leads to
>> > > timeout.
>> > > >>
>> > > >>
>> > > >> Tsz-Wo
>> > > >>
>> > > >>
>> > > >> On Fri, May 31, 2024 at 8:55 AM Ashish <[email protected]>
>> > wrote:
>> > > >>
>> > > >>> +1 (non-binding)
>> > > >>>
>> > > >>> - Ran the build on Mac/JDK 11, build fine and all test cases pass
>> > > >>>
>> > > >>> Side note, on the 1st try I did get test failure due to timeout.
>> > > >>> However, was not reproduced, shall keep an eye on the failure, in
>> > case
>> > > it
>> > > >>> happens again.
>> > > >>>
>> > > >>> Thanks
>> > > >>> Ashish
>> > > >>>
>> > > >>> On Thu, May 30, 2024 at 12:38 PM Attila Doroszlai <
>> > > [email protected]>
>> > > >>> wrote:
>> > > >>>
>> > > >>>>> I’m calling a vote For Apache Ratis Release 3.1.0 rc0.
>> > > >>>>
>> > > >>>> +1
>> > > >>>>
>> > > >>>> * Verified checksum, signature
>> > > >>>> * Compared source tarball to repo at the given tag
>> > > >>>> * Ran Rat check on source
>> > > >>>> * Built from source
>> > > >>>> * Ran tests locally
>> > > >>>> * Ran regular Ozone CI using the staged jars [1] (passed, one
>> unit
>> > > >>>> test needs to be tweaked to pass)
>> > > >>>>
>> > > >>>>> The git commit hash:
>> > > >>>>> 0a34940149aef2ed6597564ca1eb5176ccd493c7
>> > > >>>>
>> > > >>>> Note: this commit hash is for 2.5.0-rc0. The right hash for
>> > 3.1.0-rc0
>> > > >>>> is bfd029cf45bf8229ef12d1d84e3692115231fb1b.
>> > > >>>>
>> > > >>>> Thanks William for preparing rc0.
>> > > >>>>
>> > > >>>> -Attila
>> > > >>>>
>> > > >>>> [1] https://github.com/adoroszlai/ozone/actions/runs/9306740993
>> > > >>>>
>> > > >>>
>> > > >>>
>> > > >>> --
>> > > >>> thanks
>> > > >>> ashish
>> > > >>>
>> > > >>
>> > >
>> > >
>> >
>>
>>
>> --
>> thanks
>> ashish
>>
>
