[GitHub] [kafka] jolshan commented on pull request #13798: KAFKA-15028: AddPartitionsToTxnManager metrics
jolshan commented on PR #13798: URL: https://github.com/apache/kafka/pull/13798#issuecomment-1611687116 Thanks for the review 😄 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [kafka] jolshan commented on pull request #13798: KAFKA-15028: AddPartitionsToTxnManager metrics
jolshan commented on PR #13798: URL: https://github.com/apache/kafka/pull/13798#issuecomment-1610267331 I ran some tests with producer-perf. I didn't see noticeable differences, but the tests were not particularly long. I can run more if needed. The `--transaction-duration` argument tells how many milliseconds before we call commit. Also note that really only one partition was produced to, so only that one verification per transaction will trigger the metric. ``` bin/kafka-producer-perf-test.sh --transaction-duration-ms 1000 --record-size 1000 --throughput -1 --num-records 100 --topic test-topic --producer.config config/producer.properties TRUNK 100 records sent, 215842.866393 records/sec (205.84 MB/sec), 16.21 ms avg latency, 153.00 ms max latency, 0 ms 50th, 119 ms 95th, 147 ms 99th, 152 ms 99.9th. 100 records sent, 223513.634332 records/sec (213.16 MB/sec), 19.28 ms avg latency, 139.00 ms max latency, 1 ms 50th, 118 ms 95th, 133 ms 99th, 138 ms 99.9th. 100 records sent, 214638.334407 records/sec (204.70 MB/sec), 19.43 ms avg latency, 134.00 ms max latency, 1 ms 50th, 123 ms 95th, 131 ms 99th, 133 ms 99.9th. KAFKA-15028
 100 records sent, 217485.863419 records/sec (207.41 MB/sec), 17.88 ms avg latency, 151.00 ms max latency, 0 ms 50th, 128 ms 95th, 145 ms 99th, 149 ms 99.9th.
 100 records sent, 229568.411387 records/sec (218.93 MB/sec), 17.15 ms avg latency, 137.00 ms max latency, 0 ms 50th, 118 ms 95th, 130 ms 99th, 136 ms 99.9th.
 100 records sent, 220653.133274 records/sec (210.43 MB/sec), 16.47 ms avg latency, 134.00 ms max latency, 1 ms 50th, 116 ms 95th, 128 ms 99th, 133 ms 99.9th. bin/kafka-producer-perf-test.sh --transaction-duration-ms 300 --record-size 1000 --throughput -1 --num-records 100 --topic test-topic --producer.config config/producer.properties TRUNK 100 records sent, 213812.272824 records/sec (203.91 MB/sec), 15.79 ms avg latency, 142.00 ms max latency, 1 ms 50th, 101 ms 95th, 136 ms 99th, 142 ms 99.9th. 
100 records sent, 213174.163291 records/sec (203.30 MB/sec), 13.00 ms avg latency, 121.00 ms max latency, 1 ms 50th, 96 ms 95th, 118 ms 99th, 120 ms 99.9th.
 100 records sent, 225580.870742 records/sec (215.13 MB/sec), 12.45 ms avg latency, 128.00 ms max latency, 1 ms 50th, 101 ms 95th, 123 ms 99th, 127 ms 99.9th. KAFKA-15028 100 records sent, 218531.468531 records/sec (208.41 MB/sec), 11.97 ms avg latency, 69.00 ms max latency, 1 ms 50th, 52 ms 95th, 65 ms 99th, 68 ms 99.9th.
 100 records sent, 217864.923747 records/sec (207.77 MB/sec), 13.21 ms avg latency, 119.00 ms max latency, 1 ms 50th, 103 ms 95th, 117 ms 99th, 119 ms 99.9th.
 100 records sent, 214868.929953 records/sec (204.91 MB/sec), 13.07 ms avg latency, 118.00 ms max latency, 1 ms 50th, 94 ms 95th, 115 ms 99th, 117 ms 99.9th. ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [kafka] jolshan commented on pull request #13798: KAFKA-15028: AddPartitionsToTxnManager metrics
jolshan commented on PR #13798: URL: https://github.com/apache/kafka/pull/13798#issuecomment-1610196155 Thanks for sharing this flame graph. I see that the histogram takes up the majority of the processCompletedSends, but compared to the total cpu usage, it's about 1%. I still think it is worth considering though, so I will look at perf and consider the debug level metric. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [kafka] jolshan commented on pull request #13798: KAFKA-15028: AddPartitionsToTxnManager metrics
jolshan commented on PR #13798: URL: https://github.com/apache/kafka/pull/13798#issuecomment-1609874663 > Although one histogram calculation here should be ok, but it would be nice if you get some producer-perf.sh data in as well to ensure that this metric isn't adversely impacting latency. Thoughts? I can take a look at this. Is there an alternative metric that you suggest if this does affect performance too much? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [kafka] jolshan commented on pull request #13798: KAFKA-15028: AddPartitionsToTxnManager metrics
jolshan commented on PR #13798: URL: https://github.com/apache/kafka/pull/13798#issuecomment-1607881974 The last builds are looking better. @divijvaidya please take a look when you get a chance. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [kafka] jolshan commented on pull request #13798: KAFKA-15028: AddPartitionsToTxnManager metrics
jolshan commented on PR #13798: URL: https://github.com/apache/kafka/pull/13798#issuecomment-1603056522 I've updated the PR, but I'm seeing a thread leak in the tests, not sure if that's just from trunk or my change. Will investigate. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [kafka] jolshan commented on pull request #13798: KAFKA-15028: AddPartitionsToTxnManager metrics
jolshan commented on PR #13798: URL: https://github.com/apache/kafka/pull/13798#issuecomment-1590237911 unit.kafka.server.AddPartitionsToTxnManagerTest.testAddPartitionsToTxnManagerMetrics() is failing. I was a bit worried this could be flaky. Will investigate. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org