Upgrade gradle, fix test logging and int test CPU usage #1235
Conversation
|
We can close #1031 once current PR is merged. |
Codecov Report
@@ Coverage Diff @@
## master #1235 +/- ##
============================================
+ Coverage 83.62% 88.63% +5.01%
- Complexity 57 61 +4
============================================
Files 6 6
Lines 348 352 +4
Branches 38 37 -1
============================================
+ Hits 291 312 +21
+ Misses 45 29 -16
+ Partials 12 11 -1
Continue to review full report at Codecov.
|
ambry-network/src/main/java/com.github.ambry.network/BlockingChannelInfo.java
Outdated
Show resolved
Hide resolved
ambry-network/src/main/java/com.github.ambry.network/BlockingChannelInfo.java
Outdated
Show resolved
Hide resolved
| @@ -793,9 +796,6 @@ private void getEntriesSinceTest(NavigableMap<MockId, NavigableSet<IndexValue>> | |||
| highestIdIncluded = referenceIndex.lastKey().equals(idToCheck) ? null : referenceIndex.lastKey(); | |||
| // check the case where maxSize is more than the number of entries in the segment | |||
| doGetEntriesSinceTest(referenceIndex, segment, idToCheck, maxSize + 1, 0, highestIdIncluded); | |||
There was a problem hiding this comment.
One thing we can improve in the future is doGetEntriesSinceTest is invoked multiple times and some are in the loop. I find that doGetEntriesSinceTest test both getEntriesSince() and getIndexEntriesSince() methods. Note that getIndexEntriesSince() with oneEntryPerKey = true is actually tested in getEntriesSince(). It seems we don't have to explicitly test oneEntryPerKey = true for getIndexEntriesSince() again. (My two cents.)
There was a problem hiding this comment.
For now I will leave the test since whether or not it is called is dependent on the implementation of getEntriesSince and getIndexEntriesSince is still exposed to the package. I'll think about how to further reduce the test length in future PRs.
- Upgrade to gradle 5.2.1. As part of this, I have separated the integration tests into their own task so that they can be run independently. The "allTest" target will do what the "test" target used to do. - Suppress some especially noisy logs in integration tests. These log messages were firing continuously when the node-down scenarios were tested. - Configure MockCluster replication threads to do a 100 ms sleep per iteration. This greatly reduces the CPU requirements for running the integration tests and allows us to run them on travis-ci. - Set up travis to run integration tests too.
cgtz
force-pushed
the
gradle-test-fixes
branch
from
cbac5c4 to
6c99703
Compare
integration tests into their own task so that they can be run
independently. The "allTest" target will do what the "test" target
used to do.
messages were firing continuously when the node-down scenarios were
tested.
iteration. This greatly reduces the CPU requirements for running the
integration tests and allows us to run them on travis-ci.