My 2 cents.. should the test runners have profiles like "ASF build" vs "EC2 large m/c" or something, from which the appropriate timeouts are derived, and for ASF timeouts are longer than for custom envs? Or that would make the whole test infra less trustworthy?
-Mikhail 2014-04-25 15:47 GMT-07:00 Ted Yu <[email protected]>: > Looking at https://builds.apache.org/job/hbase-0.98/ , there were 9 failed > builds out of the last 17 builds. > The success rate for > https://builds.apache.org/job/HBase-0.98-on-Hadoop-1.1/was even lower. > > I think effort of making the builds, especially hbase-0.98, more stable > should be considered. > > My two cents. > > > On Fri, Apr 25, 2014 at 3:13 PM, Andrew Purtell <[email protected]> > wrote: > > > Do we keep filing the "TestFoo occasionally fails on builds.apache.org > " > > type of issues as builds.apache.org gets slower and slower? We can see > the > > build results independent of JIRA so for documentary purposes the > rationale > > seems light. > > > > I run the 0.98 unit test suite 20 times daily on JDK 6 and 7 boxes and > have > > not observed failures or zombies for a while now. Those EC2 VMs are > clearly > > reasonable test environments compared to builds.apache.org, sadly. I'm > > tempted to close any test issue reporting something on > > builds.apache.orgthat I don't see as Cannot Reproduce but wonder how > > common that feeling is. > > > > Of course small patches to increase a timeout here or retry more often > > there could be useful and acceptable. At the same time, do we increase > the > > tolerances for builds.apache.org and trade away the effectiveness of the > > test to catch real timing issues? > > > > > > -- > > Best regards, > > > > - Andy > > > > Problems worthy of attack prove their worth by hitting back. - Piet Hein > > (via Tom White) > > > -- Thanks, Michael Antonov
