On Tue, 21 Nov 2023 13:01:55 GMT, Kevin Walls <kev...@openjdk.org> wrote:
> This test fails occasionally, and can be tuned to avoid most of the failures > ever seen. > > It exists to check that notification work does not scale linearly with number > of MBeans, so it calculates a ratio of the time taken with one, and with > 20,000 MBeans. > > We should increase point at which that calculated ratio is considered a > failure. I chose 2500 here as it solves most of the failures that I think > have ever been reported with the test, without being too forgiving that we > ignore a serious slowdown in future. Looks okay and trivial to me. ------------- Marked as reviewed by sspitsyn (Reviewer). PR Review: https://git.openjdk.org/jdk/pull/16761#pullrequestreview-1751200573