Closed velo closed 8 years ago
@mpollmeier hey, do you have the means to summon luca here?
On travis it fails with an OutOfMemory error, or is there anything else I missed?
Running org.apache.tinkerpop.gremlin.orientdb.gremlintest.structure.OrientGraphStructureStandardIT
Exception in thread "Thread-8" java.lang.OutOfMemoryError: Java heap space
That's something we should be able to fix, probably with a change in .travis.yml
?
I just ran mvn install
locally and only get one error which looks easy to fix:
searchMatching(org.apache.tinkerpop.gremlin.orientdb.MatcherStepTest) Time elapsed: 0.569 sec <<< FAILURE!
java.lang.AssertionError:
Expected: (map containing ["a"->"josh"] and map containing ["c"->"marko"])
but: map containing ["a"->"josh"] map was [<a=peter>, <c=marko>]
Just mention luca if you have a question. He's very happy about our little driver project here and typically responds quickly.
Yes, it does... but on my machine it fails as well.... odd!
@lvca so, should we just give orient 2.2 more memory or this is something we should investigate a little big further?
I just realized that we give tests 2Gb to run
I don't know it is helpful, but I ran 'mvn install' and got same result with mpollmeier.
Tests run: 2, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.928 sec <<< FAILURE! searchMatching(org.apache.tinkerpop.gremlin.orientdb.MatcherStepTest) Time elapsed: 0.573 sec <<< FAILURE! java.lang.AssertionError: Expected: (map containing ["a"->"josh"] and map containing ["c"->"marko"]) but: map containing ["a"->"josh"] map was [<a=peter>, <c=marko>] at org.hamcrest.MatcherAssert.assertThat(MatcherAssert.java:20) ...
Ya, same OOM issue
Sorry for the delay, I'm on it.
I've added in the execution setting this new one that is needed from OrientDB v2.2: -XX:MaxDirectMemorySize=512g
, otherwise the JVM wont allows to use big space in direct memory.
Hrmmm, i don't think I have a machine capable of running it :/
512g is only an upper limit, I think 2G for the test is enough.
@lvca no sucess =/
@lvca Could you please help this work? I really really want to see this driver bumped to OrientDB 2.2!
Sorry guys, I'm clearly the bottleneck on this. I'm forwarding this to somebody else of the OrientDB team.
that would be nice @lvca
I ran out of ideas.... no matter what I do there is not enough memory!
@tglman and @maggiolo00 from OrientDB Team are looking into it.
Hi @velo
let me try to reproduce it
@velo
Do you run the tests on a Windows machine?
Yes, but the CI is Linux and also fails
On Tue, 6 Sep 2016, 22:01 Enrico Risa notifications@github.com wrote:
@velo https://github.com/velo
Do you run the tests on Windows machine?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mpollmeier/orientdb-gremlin/pull/74#issuecomment-244905590, or mute the thread https://github.com/notifications/unsubscribe-auth/AAIVjksxzkmur7qgJNCKeAed4PUFSAkUks5qnTnlgaJpZM4Ik0Nr .
Now seems stuck
let me check
@velo
i did try with 512g for Max memory and worked for me with 1 test fail but without OOM
XX:MaxDirectMemorySize=512g
can you try?
Wow, I don't think I have access to this kind of hardware.
Also, this seems to be a serious memory leak on 2.2 series. I can run 2.1 tests w/o 512g of memory
On Wed, 7 Sep 2016, 02:11 Enrico Risa notifications@github.com wrote:
@velo https://github.com/velo
i did try with 512g for Max memory and worked for me with 1 test fail but without OOM
XX:MaxDirectMemorySize=512g
can you try?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mpollmeier/orientdb-gremlin/pull/74#issuecomment-244962281, or mute the thread https://github.com/notifications/unsubscribe-auth/AAIVjn0awgQXQG236nL9O4ZMlBnX-Nzbks5qnXSMgaJpZM4Ik0Nr .
@velo
as @lvca explain, this is only an upper limit. OrientDB does not require that in order to run I did run the tests on my mac with such config
we can try it for debugging purposes (@velo please go ahead)
but I agree with @velo that it's not a good idea to actually set the limit so high. it does look like a memory leak, and if orient actually runs anywhere near that value it will slow down or even blow up. if it actually needs that much in our tests you should really take a heap dump and fix the problem.
@mpollmeier
i agree 2G should be fine, i think i may have found something, i will keep you updated
Tests passed on my computer using this humongous amount of memory... It did allocated 10Gb (o.0)
Still that failed on CI
@velo
I've made this PR that implements the graph.drop.
I've used it in the Test Suite so the storages used in tests are removed in teardown. You can set it back the -XX:MaxDirectMemorySize=2g paramenter
@mpollmeier any idea why the reconnect test is fail on linux but passing on windows?!
NVM, fails everywhere =D
@velo
you did remove entirely the -XX:MaxDirectMemorySize JVM settings.
Use it with 2g -XX:MaxDirectMemorySize=2g
There is no need, tests are passing now, thanks for the assistance!
On Thu, 8 Sep 2016, 19:44 Enrico Risa notifications@github.com wrote:
@velo https://github.com/velo
you did remove entirely the -XX:MaxDirectMemorySize JVM settings.
Use it with 2g -XX:MaxDirectMemorySize=2g
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mpollmeier/orientdb-gremlin/pull/74#issuecomment-245519122, or mute the thread https://github.com/notifications/unsubscribe-auth/AAIVjq9BdcfbE61AovAmVbOS949lTAzzks5qn7zBgaJpZM4Ik0Nr .
OrientReconnect test is failing because the graph instance behaviour changed - it's returning cached results instead of failing when the underlying connection drops. Will provide a fix shortly and merge everything to master.
just released 3.2.1.1 which upgrades to OrientDB 2.2 :)
Coverage decreased (-7.2%) to 73.909% when pulling c93dbf70db07f2f636a9a865234235c4c9b10b7d on velo:orient-2.2 into 36d80f6a609cd3285a560171b9ad77a681a0f402 on mpollmeier:master.