Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CNDB-11508 main-5.0 branch #1408

Open
wants to merge 16 commits into
base: main-5.0
Choose a base branch
from
Open

Conversation

ekaterinadimitrova2
Copy link
Collaborator

@ekaterinadimitrova2 ekaterinadimitrova2 commented Nov 7, 2024

What is the issue

We cannot:

  • compile and run main-5.0 on JDK22
  • run tests on JDK22 with the current config

What does this PR fix and why was it fixed

...

  • Adding test fixes based on testing we already did in a POC branch; enables us to build and run Cassandra on JDK22 plus run all test suites on JDK22 from the main-5.0 branch. We also added numerous fixes that get CI to a very good state on JDK22. CI runs posted on the GH issue.
  • apply a few fixes for JDK17 to bring CI to a good shape too. CI runs posted on the GH issue.

Checklist before you submit for review

  • Make sure there is a PR in the CNDB project updating the Converged Cassandra version
  • Use NoSpamLogger for log lines that may appear frequently in the logs
  • Verify test results on Butler
  • Test coverage for new/modified code is > 80%
  • Proper code formatting
  • Proper title for each commit staring with the project-issue number, like CNDB-1234
  • Each commit has a meaningful description
  • Each commit is not very long and contains related changes
  • Renames, moves and reformatting are in distinct commits

@ekaterinadimitrova2
Copy link
Collaborator Author

ekaterinadimitrova2 commented Nov 8, 2024

CI is in a bad shape today going up and down and I am not being able to really check test results.
I am pushing some review feedback updates and JDK17 test fixes and I will move the PRs to draft so they do not trigger useless CI triggers. I will move to another issue I am working on and check back later.

@@ -171,6 +171,9 @@
#-Xms4G
#-Xmx4G

# Need experimental bytebuddy for JDK21
-Dnet.bytebuddy.experimental
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we just upgrade to a newer byte buddy? JDK 22 is officially supported in 1.14.9 and the latest 1.15.10 supports up to jdk24.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually we are already using 1.14.17 which supports up to JDK 23. So we should not need this already.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I added it because our tests showed we need it -
https://jenkins-stargazer.aws.dsinternal.org/job/ds-cassandra-build/867/#showFailuresLink and https://jenkins-stargazer.aws.dsinternal.org/job/ds-cassandra-build/867/testReport/junit/org.apache.cassandra.cache/ChunkCacheInterceptingTest/tests_stage_1___jvm_unit_tests___jvm_unit_compression_tests___testChunkCacheInterception_compression_jdk22/
I believe our friend mockito-inline was to be blamed here. Tatu had the same issues with Stargate, if I remember correctly.

I believe the last I tried a few months ago the latest version did not work with mockito-inline so I had to downgrade and adjust versions in the POC branch. Maybe something has changed since than, so I suggest we open a follow up ticket for that?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

but this is jvm-server.options which is not used in unit tests, is it?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good question, I don't think we actually needed. I will remove it and see if CI complains for some reason that I cannot recall now.

build.xml Outdated
@@ -515,6 +606,7 @@
<jvmarg value="-Dstorage-config=${test.conf}"/>
<jvmarg value="-Dcassandra.reads.thresholds.coordinator.defensive_checks_enabled=true" /> <!-- enable defensive checks -->
<jvmarg value="-javaagent:${build.dir.lib}/jars/jamm-${jamm.version}.jar" />
<jvmarg value="-Dnet.bytebuddy.experimental=true"/>
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

none of the byte buddy -D should be needed as we are using version 1.14.17 which supports up to JDK23

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Check this comment:
#1408 (comment)
Issues with mockito-inline

Copy link
Member

@JeremiahDJordan JeremiahDJordan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

looks good. we should remove all the byte buddy -D, they should not be needed. possibly upgrade to the latest byte buddy to prevent needing to touch it again later.

@ekaterinadimitrova2
Copy link
Collaborator Author

ekaterinadimitrova2 commented Nov 13, 2024

Thank you for the review!

looks good. we should remove all the byte buddy -D, they should not be needed. possibly upgrade to the latest byte buddy to prevent needing to touch it again later.

Please check below comment:
#1408 (comment)

@@ -122,11 +122,16 @@ jvmver=`echo "$java_ver_output" | grep '[openjdk|java] version' | awk -F'"' 'NR=
JVM_VERSION=${jvmver%_*}
short=$(echo "${jvmver}" | cut -c1-2)

JAVA_VERSION=17
JAVA_VERSION=22
if [ "$short" = "11" ] ; then
JAVA_VERSION=11
elif [ "$JVM_VERSION" \< "17" ] ; then
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is this correct and consistent with what we have in main? I suggested to allow only LTS versions but eventually @JeremiahDJordan opposed and we allow for any >= 11; While here I can see 12..16 and 18..21 seem to be disallowed?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There is a patch in trunk that adds a flag to be able to allow those. We should port it and not allow it by default. But let's not do it here?
This comment I wrote in the issue:
UPDATE: I will pull the patch from CASSANDRA-18688 in a separate ticket. We can add it when we fix all tests and enable main branch for testing with JDK22 in CI. Otherwise, there will be confusion that the branch is JDK22 ready when it is still not fully ready.

<string>--add-opens java.base/jdk.internal.reflect=ALL-UNNAMED</string>
<string>--add-opens java.base/jdk.internal.vm=ALL-UNNAMED</string>
<string>--add-opens java.base/sun.nio.ch=ALL-UNNAMED</string>
<string>--add-opens jdk.compiler/com.sun.tools.javac=ALL-UNNAMED</string>
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have similar thoughts as for main - we should leverage the fact we mostly extend the sets of exports and opens with newer JVM versions

Copy link
Collaborator Author

@ekaterinadimitrova2 ekaterinadimitrova2 Nov 20, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I already responded on the other PR, it applies to this comment too:
#1398 (comment)

--add-opens java.base/java.nio=ALL-UNNAMED
--add-opens java.base/java.util=ALL-UNNAMED
--add-opens java.base/java.util.concurrent.atomic=ALL-UNNAMED
--add-opens java.base/sun.nio.ch=ALL-UNNAMED
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Similarly as in main there are duplications

build.xml Outdated
@@ -1247,6 +1340,7 @@
<jvmarg value="-Dcassandra.ring_delay_ms=1000"/>
<jvmarg value="-Dcassandra.tolerate_sstable_size=true"/>
<jvmarg value="-Dcassandra.skip_sync=true" />
<jvmarg value="-Dnet.bytebuddy.experimental=true"/>
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think you need to add it individually to each test configuration, to IntellliJ files, etc. - just add it to jdk22 props which are applied everywhere

@@ -790,7 +790,8 @@ TrieEntry<K, V> nextEntryImpl(TrieEntry<K, V> start, TrieEntry<K, V> previous, T
* This is implemented by going always to the left until
* we encounter a valid uplink. That uplink is the first key.
*/
TrieEntry<K, V> firstEntry()
// @Override needed in JDK 21+.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if @Override is needed, then why don't you add it?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Because it is needed only if we build on JDK21+. There is nothing to override on previous version and we still build on 11.

@ekaterinadimitrova2
Copy link
Collaborator Author

Feedback addressed. @jacek-lewandowski , I believe the main question here is:

let me know what you want me to do with this one - #1398 (comment)
Option 1: remove it from both main and main-5.0 and diverge from Apache (I am against that);
Option 2: remove the deletions in main and open follow up ticket to improve that target for all branches? (I prefer this one)

* CNDB-11498: Compile on JDK22:
- firstEntry() and lastEntry() were trying to assign weaker access privileges (e.g., private or package-private) when they need to be public according to the interface they are implementing (SequencedMap).
- JDK 21+ adds SequencedCollection:
The List and NavigableSet interfaces now inherit from a new interface called SequencedCollection, which introduces the reversed() method.
However, NavigableSet.reversed() returns a NavigableSet, while List.reversed() returns a List. This results in a conflict, as the types are incompatible.
…roperties `java.default` and `java.supported`.

ant generate-idea-files now support JDK 8, JDK 11 and JDK 22.
To add support of another JDK the java-jvmargs property must be set for the JDK in question (see how it's done in build.xml for Java 11 and 22)

Other minor, but notable changes are:
- test jvmargs are now added to idea run configurations
- .idea dir and project iml file are first removed and then recreated during `ant generate-idea-files`

Based on what was done in CASSANDRA-18467, CASSANDRA-18179, CASSANDRA-18258 for 17 plus additional stuff for 21

Co-authored-by: Ekaterina Dimitrova<[email protected]>
Co-authored-by: Mick Semb Wever <[email protected]>
Co-authored-by: Jakub Zytka <[email protected]>

Fix add-opens and add-exports

CNDB-11508: CQLSHLIB tests update

Fix EmptyValuesTest

Fix sjk

Fix tools/bin/cassandra.in.sh; fix some JDK internals openings for audit logger and vectorizedMismatch

Upgrade Mockito to fix UnifiedCompactionStrategyTest

Get rid of JDK8 in scripts, not supported anywhere anymore
…aybe we still want to check newer version for JDK22 specifically.

Though this is the last ecj version to support JDK11.
Upgrade:
- ecj plus fix the java udf functions for JDK11+
- snakeyaml - it was already bumped in CNDB for security vulnerability
- test dependencies:
   jacoco, byteman - higher version than CNDB but it is needed for catching up on JDK22 in tests
   findbugs - aligned with CNDB version but we probably want at some point to get to major version upgrade; not a priority for now
   jmh, bytebuddy - bumped to latest versions as they are known for not working on newer JDK versions
When running tests with JDK17 we see NoClassDefFoundError when trying
to instante org.apache.cassandra.Util class due caused by
getSupportedMTimeGranularity trying to access a private field.
This patch modified the build configuration to add a jvm option
allowing that access.
@ekaterinadimitrova2
Copy link
Collaborator Author

ekaterinadimitrova2 commented Nov 22, 2024

Thank you for the reviews! There are a few conversations about how we can improve/change certain things like generate-idea-files ant target or test JDK properties - we agreed in Slack we can tackle any of that in follow-up tickets if we want to improve things further, but the current work is already good enough to be committed as-is. I just pushed the rebased branch. To save resources, I will wait to confirm the JDK 11 PR gate looks good, and then I will push the JDK17 and JDK22 last runs and whatever else we need if there are no new issues.

@ekaterinadimitrova2
Copy link
Collaborator Author

JDK11 CI finished successfully.
There are a few test failures marked new. They all do not seem related, and all but two I cannot reproduce, seems like flakies.
The two I can reproduce, I can reproduce also on main-5.0 . They are part of - ClientRequestMetricsLatenciesTest

Moving forward to the additional testing I will have to push manually. I publish links on the main issue.

Copy link

sonarcloud bot commented Nov 22, 2024

@cassci-bot
Copy link

❌ Build ds-cassandra-pr-gate/PR-1408 rejected by Butler


18 new test failure(s) in 6 builds
See build details here


Found 18 new test failures

Showing only first 15 new test failures

Test Explanation Branch history Upstream history
t.TestCqlshUnicode.test_unicode_desc flaky 🔴🔵🔴🔵🔴🔴
t.TestCqlshUnicode.test_unicode_identifier flaky 🔴🔵🔴🔴🔴🔵
t.TestCqlshUnicode.test_unicode_value_round_trip flaky 🔵🔴🔵🔴🔵🔴
...adCommitLogAndSSTablesWithDroppedColumnTestCC40 failing 🔴🔴🔴🔴🔴🔴
...adCommitLogAndSSTablesWithDroppedColumnTestCC50 failing 🔴🔴🔴🔴🔴🔴
...oadCommitLogAndSSTablesWithDroppedColumnTestDSE failing 🔴🔴🔴🔴🔴🔴
...thRestartTest.testReadingValuesOfDroppedColumns failing 🔴🔴🔴🔴🔴🔴
...rrorsTest.testNoSuchRepairSessionAnticompaction flaky 🔴🔵🔴🔴🔴🔴
o.a.c.d.t.r.ForceRepairTest.forceWithDifference regression 🔴🔴🔵🔴🔴🔵
...ereUpdateMakesRowMatchNonOrderingPredicates[dc] regression 🔴🔵🔵🔵
o.a.c.i.c.CompressionMetadataTest.testMemoryIsF... failing 🔴🔴🔴🔴🔴🔴
....i.c.CompressionMetadataTest.testMemoryIsShared failing 🔴🔴🔴🔴🔴🔴
...ientRequestMetricsLatenciesTest.testReadMetrics regression 🔴🔴🔵🔵
...entRequestMetricsLatenciesTest.testWriteMetrics regression 🔴🔴🔵🔵
o.a.c.u.b.BinLogTest.testTruncationReleasesLogS... regression 🔴🔴🔵🔵

Found 3 known test failures

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

7 participants