The bug was identified when I looked into why a try-job has to be triggered for the first data point as in this flake analysis https://17298-f45284b-dot-frontend-dot-findit-for-me.appspot.com/waterfall/flake?key=ag9zfmZpbmRpdC1mb3ItbWVyyQELEhdNYXN0ZXJGbGFrZUFuYWx5c2lzUm9vdCKSAWNocm9taXVtLndpbi9XaW43IFRlc3RzIChkYmcpKDEpLzcxMDg0L2Jhc2VfdW5pdHRlc3RzL1VHeGhkR1p2Y20xVWFISmxZV1JVWlhOMExsTmxkRU4xY25KbGJuUlVhSEpsWVdSUWNtbHZjbWwwZVZkcGRHaFVhSEpsWVdSTmIyUmxRbUZqYTJkeWIzVnVaQT09DAsSE01hc3RlckZsYWtlQW5hbHlzaXMYAQw https://pantheon.corp.google.com/datastore/entities/query?project=findit-for-me&ns=&kind=IsolatedTarget&filter=12%2Fbuilder_name%7CSTR%7CEQ%7C17%2FWin%20Builder%20(dbg)&filter=11%2Ftarget_name%7CSTR%7CEQ%7C14%2Fbase_unittests&filter=15%2Fcommit_position%7CINT%7CEQ%7C6%2F586600 Example builds: https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Win%20Builder%20%28dbg%29/51404
For this builder, there were 44 builds between (including) these two builds. However, only 11 builds were indexed for base_unittests while 44 builds were indexed for aura_unittests. https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Win%20Builder%20%28dbg%29/51447 https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Win%20Builder%20%28dbg%29/51404 https://pantheon.corp.google.com/datastore/entities/query?project=findit-for-me&ns=&kind=IsolatedTarget&filter=12%2Fbuilder_name%7CSTR%7CEQ%7C17%2FWin%20Builder%20(dbg)&filter=11%2Ftarget_name%7CSTR%7CEQ%7C14%2Fbase_unittests&filter=12%2Fgitiles_host%7CSTR%7CEQ%7C25%2Fchromium.googlesource.com&filter=15%2Fgitiles_project%7CSTR%7CEQ%7C12%2Fchromium%2Fsrc&filter=11%2Fgitiles_ref%7CSTR%7CEQ%7C17%2Frefs%2Fheads%2Fmaster&filter=9%2Fhas_patch%7CBL%7CEQ%7C5%2Ffalse&filter=11%2Fmaster_name%7CSTR%7CEQ%7C12%2Fchromium.win&filter=15%2Fcommit_position%7CINT%7CGE%7C6%2F586600&filter=15%2Fcommit_position%7CINT%7CLE%7C6%2F586854&sortDir=DESCENDING&sortCol=commit_position
The following revision refers to this bug: https://chromium.googlesource.com/infra/infra/+/4aeeeb1fc67a2b6de47e1e4df13c50bf2253871f commit 4aeeeb1fc67a2b6de47e1e4df13c50bf2253871f Author: Roberto Carrillo <robertocn@google.com> Date: Wed Aug 29 00:15:49 2018 [Findit] Key IsolatedTarget by build_id/target/isolate_hash. Instead of the previous isolate_hash-only key. This allows us to record entries for builds that have the same isolated hash as older builds. Which is expected due to incremental compilation. This change replaces the isolated_hash property of the model to use the new column when available and fall back to the key, so that it works with the old records as well. Also remove the .Get method which isn't used except for tests and replace its use with .get_by_id included as part of ndb.Model. R=stgao,chanli,lijeffrey,liaoyuke Bug: 878552 Change-Id: Ib8dd3f5c42d513283152bc49aca22350e47fe0bc Reviewed-on: https://chromium-review.googlesource.com/1194893 Commit-Queue: Roberto Carrillo <robertocn@chromium.org> Reviewed-by: Shuotao Gao <stgao@chromium.org> Reviewed-by: Chan Li <chanli@chromium.org> [modify] https://crrev.com/4aeeeb1fc67a2b6de47e1e4df13c50bf2253871f/appengine/findit/model/test/isolated_target_test.py [modify] https://crrev.com/4aeeeb1fc67a2b6de47e1e4df13c50bf2253871f/appengine/findit/model/isolated_target.py [modify] https://crrev.com/4aeeeb1fc67a2b6de47e1e4df13c50bf2253871f/appengine/findit/handlers/test/completed_build_pubsub_ingestor_test.py [modify] https://crrev.com/4aeeeb1fc67a2b6de47e1e4df13c50bf2253871f/appengine/findit/pipelines/flake_failure/get_isolate_sha_pipeline.py
Comment 1 by st...@chromium.org
, Aug 28