kevinrr888 commented on PR #4619: URL: https://github.com/apache/accumulo/pull/4619#issuecomment-2140845122
I decreased the number of txns from 500 to 250, for me, this decreased the time to delete the table from 10-15 seconds down to 2-3 seconds. Hoping this speedup is enough for Jenkins to run this test successfully. For Ed, 500 txns caused a timeout, and 250 did not, so hoping Jenkins is similar. Probably shouldn't go much lower than 250 txns since less may not give enough attempts at running the print cmd before all the transactions complete. I have only ever seen it run at most 2 iterations of the do-while loop and I have run it quite a bit, but it is possible for more iterations. I tried running the test with 100 txns, and this would occasionally give me a failure when that 2nd iteration occurred (they would all be completed by then). Honestly, 250 txns would probably fail if 3 or more attempts had to occur but I think this is rare enough where it is okay. For example, on a run where 250 txns are created: 184 were still in progress at the first print attempt (which did not see a t xn complete mid print), then 77 were in progress at the second attempt which succeeded. If the second attempt didn't see a txn complete mid print, all of the txns would have probably completed and the test would have failed. Out of my last 10 runs, two iterations only occurred once so a third attempt is probably around a 1% chance. This is originally why I went with 500 transactions. I don't know if this chance of failure would be considered too flaky to have for a test. TLDR: Using 250 transactions instead of 500 has a higher chance of failure probably around a 1% chance. Is this too flaky for a test? -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
