Uh oh!
There was an error while loading. Please reload this page.
[SPARK-11198][STREAMING][KINESIS] Support de-aggregation of records during recovery - #9403
[SPARK-11198][STREAMING][KINESIS] Support de-aggregation of records during recovery#9403brkyvz wants to merge 12 commits into
Conversation
SparkQA
commented
Nov 2, 2015
Test build #44784 has finished for PR 9403 at commit
|
brkyvz
commented
Nov 2, 2015
Tested that this patch successfully de-aggregates in recovery as well. |
SparkQA
commented
Nov 3, 2015
Test build #44941 has finished for PR 9403 at commit
|
SparkQA
commented
Nov 3, 2015
Test build #44944 has finished for PR 9403 at commit
|
SparkQA
commented
Nov 3, 2015
Test build #44956 has finished for PR 9403 at commit
|
SparkQA
commented
Nov 4, 2015
Test build #44975 has finished for PR 9403 at commit
|
tdas
commented
Nov 4, 2015
test this again. |
There was a problem hiding this comment.
Isn it easier to have a parameter in the constructor? Less code while subclassing class WithAggregationKinesisStreamSuite extends KinesisStreamTests(aggregateTestData = true)
tdas
commented
Nov 4, 2015
overall the code looks fine, but Kinesis tests are not passing |
SparkQA
commented
Nov 5, 2015
Test build #45049 has finished for PR 9403 at commit
|
SparkQA
commented
Nov 6, 2015
Test build #45149 has finished for PR 9403 at commit
|
brkyvz
commented
Nov 6, 2015
test this please |
zsxwing
commented
Nov 6, 2015
The failure may be because protobuf 2.6.1 was not enabled. BTW, if 2.6.1 is enabled, shouldn't protobuf 2.6.1 break other stuff in Spark and make the end-to-end test fail? |
brkyvz
commented
Nov 6, 2015
@zsxwing probably not. Locally tests pass. They fail on jenkins for some reason. I've enabled a lot of the logging to look deeper into it. |
SparkQA
commented
Nov 6, 2015
Test build #45181 has finished for PR 9403 at commit
|
SparkQA
commented
Nov 6, 2015
Test build #45249 has started for PR 9403 at commit |
brkyvz
commented
Nov 6, 2015
test this please |
SparkQA
commented
Nov 6, 2015
Test build #45255 has finished for PR 9403 at commit
|
SparkQA
commented
Nov 9, 2015
Test build #45343 has finished for PR 9403 at commit
|
brkyvz
commented
Nov 9, 2015
Huh. Didn't change much but the tests passed this time. I wonder if it was a Java 7 vs. 8 mismatch... Just to be sure, will re-run tests |
brkyvz
commented
Nov 9, 2015
test this please |
SparkQA
commented
Nov 9, 2015
Test build #45377 has finished for PR 9403 at commit
|
brkyvz
commented
Nov 9, 2015
Aww man. The Kinesis tests passed, hive tests failed. Re-running |
brkyvz
commented
Nov 9, 2015
test this please |
SparkQA
commented
Nov 9, 2015
Test build #45391 has finished for PR 9403 at commit
|
brkyvz
commented
Nov 9, 2015
test this please |
1 similar comment
brkyvz
commented
Nov 9, 2015
test this please |
SparkQA
commented
Nov 10, 2015
Test build #45422 has finished for PR 9403 at commit
|
tdas
commented
Nov 10, 2015
LGTM. Merging this to master and 1.6. Thanks @brkyvz |
…uring recovery While the KCL handles de-aggregation during the regular operation, during recovery we use the lower level api, and therefore need to de-aggregate the records. tdas Testing is an issue, we need protobuf magic to do the aggregated records. Maybe we could depend on KPL for tests? Author: Burak Yavuz <brkyvz@gmail.com> Closes#9403 from brkyvz/kinesis-deaggregation. (cherry picked from commit 26062d2) Signed-off-by: Tathagata Das <tathagata.das1565@gmail.com>
SparkQA
commented
Nov 10, 2015
Test build #45440 has finished for PR 9403 at commit
|
lordnynex
commented
Nov 17, 2015
Despite being closed, I'm confused about why this causes hive unit tests to fail. Is this unrelated?Also, the publisher in the unit test looks like it's publishing an intentionally non-aggregated stream. Is that expected? Am I reading this incorrectly? |
brkyvz
commented
Nov 18, 2015
Hi @lordnynex. Why did you think this causes Hive tests to fail? That was most probably a flaky test. |
zsxwing
commented
Nov 30, 2015
KinesisStreamTests in test.py is broken because of this PR. See https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/46896/testReport/(root)/KinesisStreamTests/test_kinesis_stream/ Is the new dependency the failure cause? The PR builds for this PR actually didn't report the Python failure because of #9669. |
KinesisStreamTests in test.py is broken because of #9403. See https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/46896/testReport/(root)/KinesisStreamTests/test_kinesis_stream/ Because Streaming Python didn’t work when merging #9403, the PR build didn’t report the Python test failure actually. This PR just disabled the test to unblock #10039 Author: Shixiong Zhu <shixiong@databricks.com> Closes#10047 from zsxwing/disable-python-kinesis-test.
While the KCL handles de-aggregation during the regular operation, during recovery we use the lower level api, and therefore need to de-aggregate the records.
@tdas Testing is an issue, we need protobuf magic to do the aggregated records. Maybe we could depend on KPL for tests?