TEZ-4420: Upgrade to Hadoop 3.3.4 by abstractdog · Pull Request #213 · apache/tez

abstractdog · 2022-05-31T10:18:54Z

No description provided.

pjfanning · 2022-07-11T10:44:47Z

The test issues seem to be related to mockito

java.lang.NoClassDefFoundError: org/mockito/stubbing/Answer
at org.apache.tez.test.TestAM.setup(TestAM.java:66)
Caused by: java.lang.ClassNotFoundException: org.mockito.stubbing.Answer
at org.apache.tez.test.TestAM.setup(TestAM.java:66)

ayushtkn · 2022-07-11T10:56:03Z

You are checking the wrong build result, mockito also was a problem, but we upgraded it and the build result got sorted. Check the result here of this test in the latest build:
https://ci-hadoop.apache.org/job/tez-multibranch/job/PR-213/4/testReport/org.apache.tez.history/TestHistoryParser/testParserWithFailedJob/

There is an exception:

java.lang.AbstractMethodError: javax.ws.rs.core.UriBuilder.uri(Ljava/lang/String;)Ljavax/ws/rs/core/UriBuilder;
at javax.ws.rs.core.UriBuilder.fromUri(UriBuilder.java:96)
at com.sun.jersey.spi.container.servlet.ServletContainer.doFilter(ServletContainer.java:911)

steveloughran · 2022-08-04T12:55:21Z

hadoop 3.3.4 cuts jax.rs from the dependency graph

ayushtkn

We have now conflicts with the master branch. Can try with 3.3.4, which is released now to see if things work here now

zhangbutao · 2022-09-08T06:56:56Z

@abstractdog Can we upgrade hadoop to 3.3.4? I have tested hive maser with hadoop3.3.4(apache/hive#3578) , and if i change tez's hadoop dependency to 3.3.4, everything looks ok.

abstractdog · 2022-09-08T15:57:42Z

yes, let me check this PR and have some test coverage

ayushtkn

We got a clean build, the Jackson stuff changes in hadoop, got us sorted. :-)
LGTM

zhangbutao · 2022-09-13T14:54:33Z

Great! If this PR is merged into Tez 0.10.2, then we can continue to upgrade hadoop version in hive repo.

pjfanning · 2022-10-22T00:45:29Z

Can this be merged to unblock other upgrades?

abstractdog · 2022-10-22T10:29:33Z

Can this be merged to unblock other upgrades?

let me take a look soon
what upgrade is this blocking at the moment? actually, hive is not blocked now AFAIK

pjfanning · 2022-10-22T11:42:51Z

It was Hive that I thought was blocked. It would still be nice to get this moved on but I guess if it's not blocking anything then it's not as urgent.

steveloughran · 2022-10-24T10:30:37Z

we are nearing code freeze for 3.3.5 btw; if there is some final blocker to upgrade, now is the time to identify it

amanraj2520 · 2022-11-17T13:42:46Z

Hi @ayushtkn @abstractdog @pjfanning @steveloughran

Since this PR has been stale for a month, just following up to check whether I can help in validating any other test scenarios for the upgrade. Or if not, are we good to merge this PR ?

amanraj2520 · 2022-11-21T12:06:49Z

Hi @ayushtkn @abstractdog @pjfanning @steveloughran,

I am driving the hive-3.2.0 release in Open Source for which we have decided to upgrade to hadoop 3.3.4. But since Tez 0.10.2 is using hadoop-3.3.1 there can be some integration issues for this stack. I know that we are currently discussing hadoop-3.3.5 release but it would take at least a couple of months to get a stable release candidate for the same which would also require a lot of testing.

This will be a blocker for the 3.2.0 release and since we have started to plan our tasks around this release, I was proposing to make a Tez release 0.10.3 where we can have this hadoop version upgraded to 3.3.4. This would insure that we can start working on the 3.2.0 without any blockers.

Please share your thoughts on the same. I am open to discussing this release further.

Thanks,
Aman.

ayushtkn · 2022-11-21T15:05:20Z

I was talking about
apache/hive#3279
here we tried till 3.3.3 and all the unit tests were passing but on actual cluster things weren't working. So, I meant apart from the unit tests, we need to test on actual clusters as well.

The failure was flagged here:
apache/hive#3279 (comment)

amanraj2520 · 2022-11-21T15:23:20Z

Hi @ayushtkn

I am familiar with this error. This happens because Tez 0.10.1 is using Hadoop 3.1.3 which in turn uses jetty 9.3* which does not have this method but when we upgrade to Hadoop 3.3.4 it uses this method. So basically I am saying that if we maintained the same version of Hadoop in Hive and Tez, this would not have been the issue. Having said this, I will make sure that the 0.10.3 is validated on clusters including unit tests Ayush.

steveloughran · 2022-11-22T16:47:33Z

3.3.4 is the one where we fixed the tez incompatibilities for you 🙂

i plan to cut the 3.3.5 rc0 this week but it's a test of the rc process (x86 and arm) rather than something we intend to bring to a vote.

it'd be good to test hive and tez with the rc to see if we have caused any regressions, as now will be the time to fix. there's been a lot of dependency updates to stop CVEs, but we've left alone the ones we know cause problems downstream (jackson is on a cve-fixed 2.12.x release)

amanraj2520 · 2022-11-22T17:20:33Z

@steveloughran So do you think having a new Tez release with 3.3.4 should help or not?

amanraj2520 · 2022-11-24T11:56:05Z

@abstractdog @steveloughran Can you please look into this?

steveloughran · 2022-11-25T11:54:53Z

3.3.4 should be good; if not try with a 3.3.5 snapshot (build yourself) and see if that fixes things.

i'd suggest going with 3.3.4 if it works, so the 3.3.5 release isn't a blocker. you can do an upgrade after

abstractdog · 2022-11-26T07:42:20Z

I believe from tez->hadoop point of view, we usually rely only on precommit testing (including lots of minicluster tests)
if something is broken from hive -> tez -> hadoop, we should report it separately
I guess we should simply go on with hadoop 3.3.4 now, and resolve this ticket

TestRecovery passed locally, I'm restarting precommit tests

can someone approve this PR? @rbalamohan , @jteagles

tez-yetus · 2022-11-26T11:04:23Z

💔 -1 overall

Vote	Subsystem	Runtime	Comment
+0 🆗	reexec	36m 14s	Docker mode activated.
		_ Prechecks _
+1 💚	dupname	0m 0s	No case conflicting files found.
+1 💚	@author	0m 0s	The patch does not contain any @author tags.
-1 ❌	test4tests	0m 0s	The patch doesn't appear to include any new or modified tests. Please justify why no new tests are needed for this patch. Also please list what manual steps were performed to verify this patch.
		_ master Compile Tests _
+0 🆗	mvndep	6m 32s	Maven dependency ordering for branch
+1 💚	mvninstall	10m 4s	master passed
+1 💚	compile	3m 12s	master passed with JDK Ubuntu-11.0.17+8-post-Ubuntu-1ubuntu220.04
+1 💚	compile	3m 2s	master passed with JDK Private Build-1.8.0_352-8u352-ga-1~20.04-b08
+1 💚	javadoc	3m 9s	master passed with JDK Ubuntu-11.0.17+8-post-Ubuntu-1ubuntu220.04
+1 💚	javadoc	2m 26s	master passed with JDK Private Build-1.8.0_352-8u352-ga-1~20.04-b08
		_ Patch Compile Tests _
+0 🆗	mvndep	1m 17s	Maven dependency ordering for patch
+1 💚	mvninstall	6m 9s	the patch passed
+1 💚	compile	4m 7s	the patch passed with JDK Ubuntu-11.0.17+8-post-Ubuntu-1ubuntu220.04
+1 💚	javac	4m 7s	the patch passed
+1 💚	compile	3m 42s	the patch passed with JDK Private Build-1.8.0_352-8u352-ga-1~20.04-b08
+1 💚	javac	3m 42s	the patch passed
+1 💚	whitespace	0m 0s	The patch has no whitespace issues.
+1 💚	xml	0m 2s	The patch has no ill-formed XML file.
+1 💚	javadoc	3m 24s	the patch passed with JDK Ubuntu-11.0.17+8-post-Ubuntu-1ubuntu220.04
+1 💚	javadoc	2m 53s	the patch passed with JDK Private Build-1.8.0_352-8u352-ga-1~20.04-b08
		_ Other Tests _
+1 💚	unit	40m 17s	tez-tests in the patch passed.
+1 💚	unit	72m 24s	root in the patch passed.
+1 💚	asflicense	1m 14s	The patch does not generate ASF License warnings.
		201m 22s

Subsystem	Report/Notes
Docker	ClientAPI=1.41 ServerAPI=1.41 base: https://ci-hadoop.apache.org/job/tez-multibranch/job/PR-213/8/artifact/out/Dockerfile
GITHUB PR	#213
JIRA Issue	TEZ-4420
Optional Tests	dupname asflicense javac javadoc unit xml compile
uname	Linux e4562a85b404 4.15.0-191-generic #202-Ubuntu SMP Thu Aug 4 01:49:29 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux
Build tool	maven
Personality	personality/tez.sh
git revision	master / `2fd7df4`
Default Java	Private Build-1.8.0_352-8u352-ga-1~20.04-b08
Multi-JDK versions	/usr/lib/jvm/java-11-openjdk-amd64:Ubuntu-11.0.17+8-post-Ubuntu-1ubuntu220.04 /usr/lib/jvm/java-8-openjdk-amd64:Private Build-1.8.0_352-8u352-ga-1~20.04-b08
Test Results	https://ci-hadoop.apache.org/job/tez-multibranch/job/PR-213/8/testReport/
Max. process+thread count	1388 (vs. ulimit of 5500)
modules	C: tez-tests . U: .
Console output	https://ci-hadoop.apache.org/job/tez-multibranch/job/PR-213/8/console
versions	git=2.25.1 maven=3.6.3
Powered by	Apache Yetus 0.12.0 https://yetus.apache.org

This message was automatically generated.

amanraj2520 · 2023-01-04T05:17:19Z

@rbalamohan @jteagles Can you please approve this PR. Also @abstractdog are we going with the tez 0.10.3 release which will have hadoop 3.3.4.

amanraj2520 · 2023-01-04T06:49:32Z

@abstractdog Upgrade to Hadoop 3.3.4 is failing tests in Tez as follows :
[INFO] Running org.apache.tez.common.TestTezCommonUtils [ERROR] Tests run: 1, Failures: 0, Errors: 1, Skipped: 0, Time elapsed: 13.61 s <<< FAILURE! - in org.apache.tez.common.TestTezCommonUtils [ERROR] org.apache.tez.common.TestTezCommonUtils Time elapsed: 13.607 s <<< ERROR! java.lang.RuntimeException: problem starting mini dfs cluster at org.apache.tez.common.TestTezCommonUtils.setup(TestTezCommonUtils.java:71) Caused by: java.io.IOException: Timed out waiting for Mini HDFS Cluster to start at org.apache.tez.common.TestTezCommonUtils.setup(TestTezCommonUtils.java:66)

It is failing with timed out waiting for Mini DFS Cluster to start.

steveloughran · 2023-01-10T07:58:49Z

Any reason in the logs for the cluster not coming up?

amanraj2520 · 2023-01-10T08:09:05Z

@steveloughran Not checked it yet. Will check and let you know

amanraj2520 · 2023-02-07T18:06:24Z

@steveloughran @abstractdog Apologizes for getting back late. Was involved in Hive test case fixes for branch-3. Seems like this issue was intermittent. Now I see this test passing, tried it 3-4 times. Attached the snippet.

We can merge this to master. +1 from my side. Tested it extensively on my local as well as on a local cluster.

amanraj2520 · 2023-02-07T18:09:07Z

Sorry added the wrong snippet for the passing test. Attached below:

amanraj2520 · 2023-02-07T18:26:26Z

I see these test failures in the tez-tests module. Currently fixing it @abstractdog @steveloughran This issue comes when PowerMock library conflicts with Mockito as described here https://stackoverflow.com/questions/71973762/java-lang-nosuchmethoderror-org-mockito-answers-getlorg-mockito-stubbing-answ

amanraj2520 · 2023-02-07T18:33:31Z

Fixed all the tests now :

@abstractdog Can you please cherry-pick this commit I did in my forked branch - 944476e this will fix the tests. I cannot push these changes to your branch.

amanraj2520 · 2023-02-07T19:01:57Z

Still there are more failures with further tests. Will fix them

amanraj2520 · 2023-02-08T07:31:04Z

Got the issue :
There was this new commit added to TestLocalMode in TEZ-4447 which basically asserted
assertEquals(VertexStatus.State.SUCCEEDED, dagClient1.getVertexStatus(SleepProcessor.SLEEP_VERTEX_NAME, null).getState());

When I debugged this tests I found this :

When it tries to get from the cachedVertexStatus (size : 0) the status of vertexName "sleep" it does not find any entry and therefore returns null. And hence, dagClient1.getVertexStatus(SleepProcessor.SLEEP_VERTEX_NAME, null) returns null and when a getState happens, it throws null pointer exception.

@abstractdog Can you please check this from your end, maybe we need to add some config to have an entry for the Sleep vertex name in cachedVertexStatus

steveloughran · 2023-02-08T10:30:14Z

where is the mockito ref coming from? as I don't see why it should be exported -if it is, then that is something to cut back on

abstractdog · 2023-02-08T15:11:52Z

@amanraj2520: thanks for your work so far, if you have the bandwidth to drive this, please open a PR and assign TEZ-4420 to yourself
we'll find a way to cross-reference PRs as there is so much context here

btw: yeah, tez 0.10.3 can contain this change, I think once this is merged and we merge some other fixes, we can go for a release (in a couple of weeks/months)

amanraj2520 · 2023-02-08T17:00:16Z

@abstractdog Sure assigning TEZ-4420 to myself

abstractdog · 2023-02-24T07:39:14Z

closing this as final PR was #272