Turns out that I had another process running on 8080, so the tests were
fine. I made a minor update nevertheless to use available ports instead
of hardcoding. We should probably consolidate in a utility across all
tests. Not something to hold the release though. Running full tests now,
but it'll take a while.
Cheers,
Hadrian
On 01/05/2015 09:29 PM, Hadrian Zbarcea wrote:
Hi Gary,
I can confirm that your changes fixed the leveldb issue.
I did a full build, and there are a couple of failures in http, I will
look at them in the morning. Looks like port port in use conflict due
to slow shutdown from previous test (I don't think I have anything
running on 8080). Best would be to use dynamic ports, anyway. Looks
like it's getting really close.
Thanks,
Hadrian
On 01/05/2015 04:28 PM, Gary Tully wrote:
It seems there are two issues - a 2s session expiry and 500ms tick
leaves too much room for skipped pings and client and server mismatch
in session expiry. In the tests the single zk server stays alive there
are no dropped connections.
One of the problems i encountered seems to be
https://issues.apache.org/jira/browse/ZOOKEEPER-1751 and an upgrade
helps reduce
the spurious session expiry.
But to get stability - moving from the default 2s session expiry to
something like 15s is necessary. Otherwise a client may not ping in
time and the server will expiry.
I note that in 5.10 - these tests were mostly excluded via the pom, so
they would only run on linux. Recently I removed osx from the profiles
that were excluded - which may explain the diff from 5.10
On 5 January 2015 at 15:32, Dejan Bosanac <[email protected]> wrote:
Cancelling this vote. Let’s try to get to the bottom of these failures.
Regards
--
Dejan Bosanac
----------------------
Red Hat, Inc.
[email protected]
Twitter: @dejanb
Blog: http://sensatic.net
ActiveMQ in Action: http://www.manning.com/snyder/
On Mon, Jan 5, 2015 at 4:11 PM, Gary Tully <[email protected]>
wrote:
-1 : let's get a clean bill of health on the test run.
org.apache.activemq.leveldb.test.ReplicatedLevelDBBrokerTest#testAMQ4837viaJMX
will fail for me locally about 60% of the time, with a zk "Client
session timed out,"
I will try and find the root cause but I think we should add an
@Ignore and not hold up the release if the failures prove elusive.
On 31 December 2014 at 19:04, James Carman
<[email protected]>
wrote:
-1
If the build does not succeed, we need to address that before
releasing.
James
On Mon, Dec 29, 2014 at 9:26 AM, Dejan Bosanac <[email protected]>
wrote:
Hi folks,
I've just cut a release candidate for the long-awaited 5.11.0
release.
This
release has
127 bug fixes and improvements.
Could you review the artifacts and vote? Especially, it would be
great
if
you could test unix shell script and make sure there's no any
regressions
on the platform you're using.
The list of resolved issues is here:
https://issues.apache.org/jira/secure/ReleaseNote.jspa?projectId=12311210&version=12324951
You can get binary distributions here:
https://repository.apache.org/content/repositories/orgapacheactivemq-1012/org/apache/activemq/apache-activemq/5.11.0/
Source archives are here:
https://repository.apache.org/content/repositories/orgapacheactivemq-1012/org/apache/activemq/activemq-parent/5.11.0/
Maven2 repository is at:
https://repository.apache.org/content/repositories/orgapacheactivemq-1012/
Source tag:
https://git-wip-us.apache.org/repos/asf?p=activemq.git;a=commit;h=48b0cf396c18216899ceb94ee57c01685104223d
Please vote to approve this release. As it's the holiday season,
I'll
let
this vote open for longer time than usual, so everybody get the
chance
to
review it.
[ ] +1 Release the binary as Apache ActiveMQ 5.11.0
[ ] -1 Veto the release (provide specific comments)
Here's my +1
Regards
--
Dejan Bosanac
----------------------
Red Hat, Inc.
[email protected]
Twitter: @dejanb
Blog: http://sensatic.net
ActiveMQ in Action: http://www.manning.com/snyder/
