https://pantsbuild.org/ logo
Powered by
#general
  • m

    many-table-69772

    07/21/2015, 4:41 PM
    will try from scratch
  • m

    many-table-69772

    07/21/2015, 4:41 PM
    but build job succeed few times before getting to this state
  • r

    rough-minister-58256

    07/21/2015, 4:42 PM
    ah, k. zombied nailguns seem to indicate that at least one your nailgun servers is dying unexpectedly - curious why; maybe you can correlate the .pids dir to the nailgun type
  • r

    rough-minister-58256

    07/21/2015, 4:43 PM
    pants too, in fact
  • m

    many-table-69772

    07/21/2015, 4:44 PM
    cat ng_BundleCreate_bundle/pid
  • m

    many-table-69772

    07/21/2015, 4:44 PM
    this one
  • r

    rough-minister-58256

    07/21/2015, 4:46 PM
    do you see any bundle failures in the prior runs before this failure occurs?
  • m

    many-table-69772

    07/21/2015, 4:46 PM
    ng_ScroogeGen_gen_scrooge/pid this one dead as well
  • r

    rough-minister-58256

    07/21/2015, 4:49 PM
    I’m guessing its trying to run something on the nailgun in subsequent runs, sees its non-responsive then tries to kill it, fails; produces what you see. I can add some better handling to the kill check for the zombie case - should be fine to ignore it since afaik zombies are dead and should get inherited by init and reaped eventually; tho would be good to understand why they’re zombied in the first place.
  • r

    rough-minister-58256

    07/21/2015, 4:50 PM
    if you end up repro’ing from scratch, lmk - will try to reproduce myself on a linux box today and try to fix up the handling
  • m

    many-table-69772

    07/21/2015, 4:51 PM
    ^there was no failures prior
  • m

    many-table-69772

    07/21/2015, 4:51 PM
    and I can’t find anything common between 2 times when it came to this state
  • m

    many-table-69772

    07/21/2015, 4:52 PM
    thanks @rough-minister-58256 ! will clean and see if I manage to reproduce it
  • r

    rough-minister-58256

    07/21/2015, 4:52 PM
    k. there may be some ng logging under .pants.d/ng/* too
  • r

    rough-minister-58256

    07/21/2015, 4:53 PM
    that should be the ng stdout/stderr
  • r

    rough-minister-58256

    07/21/2015, 4:53 PM
    may provide some clues
  • m

    many-table-69772

    07/21/2015, 5:12 PM
    Copy code
    OpenJDK 64-Bit Server VM warning: ignoring option MaxPermSize=256m; support was removed in 8.0
java.io.EOFException
	at java.io.DataInputStream.readInt(DataInputStream.java:392)
	at com.martiansoftware.nailgun.NGSession.run(NGSession.java:182)
  • m

    many-table-69772

    07/21/2015, 5:12 PM
    in each stderr (BundleCreate_bundle IvyResolve_resolve_ivy ScalaZincCompile ScroogeGen_gen_scrooge)
  • m

    many-table-69772

    07/21/2015, 5:13 PM
    that is after full cleanup and container restart
  • w

    witty-crayon-22786

    07/21/2015, 5:13 PM
    (iirc, the EOFException was an existing thing / potentiall red herring)
  • m

    many-table-69772

    07/21/2015, 5:15 PM
    and I see bunch of zombie processes after single run
  • w

    witty-crayon-22786

    07/21/2015, 5:29 PM
  • w

    witty-crayon-22786

    07/21/2015, 5:30 PM
    (twitter is getting ready to change our consume strategy, but at this instant we need pants internal APIs to match between the SHA and release we're using)
  • r

    rough-minister-58256

    07/21/2015, 5:58 PM
    yeah, even I see EOFException in mine
  • w

    witty-crayon-22786

    07/21/2015, 6:02 PM
    @ancient-stone-50795: can you help with https://rbcommons.com/s/twitter/r/2505/ ?
  • w

    witty-crayon-22786

    07/21/2015, 6:09 PM
    @enough-analyst-54434: fixed https://rbcommons.com/s/twitter/r/2505/ .. thanks!
  • w

    witty-crayon-22786

    07/21/2015, 6:22 PM
    @enough-analyst-54434, @ancient-stone-50795 : argh... sorry guys. broke CI. will wait for it to finish this time, but here's the fix: https://rbcommons.com/s/twitter/r/2507/
  • a

    adorable-hairdresser-18353

    07/21/2015, 6:31 PM
    how do people currently use the .pants.d/tests directory currently created by JUnit? I’m thinking of also having a merged result output to the run report and wanted your thoughts
  • a

    ancient-stone-50795

    07/21/2015, 6:36 PM
    As an example, @adorable-hairdresser-18353 and I already discussed that Square has an internal use of the xml outputs. We take these reports and use them to upload to our CI builder so it can track flaky tests.
  • e

    enough-analyst-54434

    07/21/2015, 6:46 PM
    Before option work / simplification there did used to exist a flag to control the output dir for the xml reports for invokers who wanted to consume them but not assume anything about pants internal dirs
1...646566...2420Latest