1. 26 Jul, 2018 2 commits
    • Harald Sitter's avatar
      properly set iso relation + disable concurrent builds · 6eeec135
      Harald Sitter authored
      calling properties in a jenkinsfile overrides whatever was set in the
      xml. this is a problem as running a job that does it will undo the
      concurrent disablement done in the tooling templates. so, we also disable
      here. tooling is the default value. the values from the jenkinsfile only
      get applied after first run
      also properly init openqa_series to xenial unless otherwise defined
      (which it should be moving forward FTR)
    • Harald Sitter's avatar
      only claim one openqa install slot · 03f1841e
      Harald Sitter authored
      this allows us to add more slots via jenkins if need be (e.g. to get
      multiple test through faster, at the risk of having flakes)
  2. 06 Jul, 2018 1 commit
  3. 05 Jul, 2018 2 commits
    • Harald Sitter's avatar
      fix syntax of previous commit · e319eb21
      Harald Sitter authored
    • Harald Sitter's avatar
      fix archiving when install tests fail early · 9222a60f
      Harald Sitter authored
      when an install test fails before it reaches the original metadata
      retrieval (e.g. broken uefi) the metadata dir may be empty, therefore
      we cannot expect it not to be empty.
      instead we'll allow empty, do the rest of the post-build archiving and
      then explicitly attempt to ls * which should then end in error if sh
      cannot expand * to anything.
      without this the metadata archival aborts in error and none of the steps
      after it are actually reached. also, moving the archival to the end made
      similiarly no sense since the steps such as chown may also fail, so we'd
      then skip archiving on account of unrelated error. archiving and split
      assertion seems the best way really
  4. 13 Jun, 2018 2 commits
    • Harald Sitter's avatar
      change locking to happen before claiming the node · 744a7da1
      Harald Sitter authored
      before we'd claim the node,, enter a stage and then lock the serializer.
      while that is fine from a serialization pov it's bogus from a node
      pov as we then sit there in the build slot of the node without doing
      so, instead claim the lock and then the node. this way we are basically
      "queued" up until we have the lock at which point we then also claim our
      build slot
    • Harald Sitter's avatar
      limit concurrent installation tests via jenkins resources · aa8a1969
      Harald Sitter authored
      we are having continued problems with installation tests semi-randomly
      failing. the working theory is that there are IO problems caused by
      concurrently running install tests using too many resources. as a quick
      fix we'll serialize the installations, which should, in theory anyway,
      reduce the number of flaky runs to zero
  5. 24 Apr, 2018 1 commit
  6. 20 Apr, 2018 1 commit
  7. 17 Apr, 2018 1 commit
  8. 14 Feb, 2018 1 commit
  9. 13 Feb, 2018 2 commits
    • Harald Sitter's avatar
      archive the wok relative · e6ae0008
      Harald Sitter authored
      previously we'd have to dig into wok to get to the wok's artifacts, since
      they are the primary artifacts that is fairly daft though. by diring
      into the wok first we can archive them into the root of the artifacts dir
    • Harald Sitter's avatar
      add a new metadata system to get high level access to global metadata · 4d94c67d
      Harald Sitter authored
      this is primarily useful to get metadata out of the ISO so we know
      what was installed initially.
      this data is uploaded by the install core as soon as possible and later
      (on success) moved to a metdata dir for archival.
      so, for installs the metadata is in ulogs on error and metadata on success
      for !install tests they pick up the metadata from the image storage to
      seed their initial metadata dir (they may still added more info if
      they always have metadata (from the install), so even on errors we know
      what the image contained
      archival of this happens relative, so the metadata is in the main dir
  10. 07 Feb, 2018 1 commit
  11. 06 Feb, 2018 1 commit
  12. 11 Jan, 2018 1 commit
  13. 10 Jan, 2018 2 commits
  14. 14 Dec, 2017 1 commit
  15. 07 Dec, 2017 1 commit
  16. 03 Jul, 2017 1 commit
  17. 23 Jun, 2017 4 commits
  18. 22 Jun, 2017 1 commit
  19. 21 Jun, 2017 2 commits
  20. 19 Jun, 2017 2 commits
  21. 16 Jun, 2017 1 commit
  22. 14 Jun, 2017 5 commits
  23. 13 Jun, 2017 4 commits