Hello Builds I'm investigating a regular test failure we (Qpid) see on one of our Jenkins jobs (Qpid-Java-Java-DBY-TestMatrix). The job is a matrix that is tied to the ubuntu label. I notice that the test failure (a timeout) seems to occur when ubuntu4 or 5 is chosen as a build slave. Tests that normally take a seconds or less extend to tens of seconds or worse and the job is caught by the build timeout plugin and a failure results.
I'm still investigating, but at the moment, the application logs are pointing to a very slow filesystem as a likely cause. I also see similar results timing sync and dd commands from a jenkins job. Script fragment and results are below. It would be helpful to me to know the spec of juno and janus machines including their filesystem spec and config (eg. RAID) (the details are absent from www.apache.org/dev/machines.html). Are these true machines or guest VMs? Are you aware of anything else that may cause slow filesystem performance? Kind regards, Keith Wall. Script: time sync sh -c 'dd if=/dev/zero of=ddfile bs=8k count=200000 && sync' time sync ubuntu5 time sync (before) real 0m6.513s user 0m0.000s sys 0m0.584s 200000+0 records in 200000+0 records out 1638400000 bytes (1.6 GB) copied, 3.58202 s, 457 MB/s real 0m32.567s user 0m0.016s sys 0m4.116s time sync (after) real 0m1.378s user 0m0.000s sys 0m0.552s ubuntu3 (for comparison): time sync (before) real 0m0.491s user 0m0.000s sys 0m0.372s 1638400000 bytes (1.6 GB) copied, 1.25777 s, 1.3 GB/s real 0m4.696s user 0m0.012s sys 0m1.608s time sync (after) real 0m0.376s user 0m0.000s sys 0m0.364s
