kahlee-paladin is slow, not using expected prebuilts. |
|||||
Issue descriptionpprabhu: What's up with Kahlee paladin? It's gotten the slowest by far. akeshet: Yes, median BuildPackages time for it has grown steadily http://shortn/_UODNEPpWFa especially in the tail. Looking at a recent example https://luci-milo.appspot.com/buildbot/chromeos/kahlee-paladin/1465# I see: 1) amd-firmware was retried, but that doesn't really explain things. Filed https://bugs.chromium.org/p/chromium/issues/detail?id=816604 anyway 2) even though there's only 1 relevant CL, I see a large number of packages that are being build from ebuild (ie being rebuilt) rather than from binary package and BuildPackages stage is pushing 50 minutes. Is something wrong with our binary package publishing? akeshet: Digging more into https://luci-milo.appspot.com/buildbot/chromeos/kahlee-paladin/1465 In it's SetupBoard stage https://logs.chromium.org/v/?s=chromeos%2Fbb%2Fchromeos%2Fkahlee-paladin%2F1465%2F%2B%2Frecipes%2Fsteps%2FSetupBoard%2F0%2Fstdout I see which binhost is is trying to use... 08:29:06: INFO: RunCommand: /mnt/host/source/.cache/common/gsutil_4.27.tar.gz/gsutil/gsutil -o 'Boto:num_retries=10' cp -v -- gs://chromeos-prebuilt/board/amd64-generic/paladin-R66-10436.0.0-rc2/packages/Packages /tmp/tmpJhvvEk.tmp 08:29:08: INFO: RunCommand: /mnt/host/source/.cache/common/gsutil_4.27.tar.gz/gsutil/gsutil -o 'Boto:num_retries=10' cp -v -- gs://chromeos-prebuilt/board/kahlee/paladin-R66-10436.0.0-rc2/packages/Packages /tmp/tmpMV2lV5.tmp 08:29:09: INFO: RunCommand: /mnt/host/source/.cache/common/gsutil_4.27.tar.gz/gsutil/gsutil -o 'Boto:num_retries=10' cp -v -- gs://chromeos-prebuilt/board/caroline/chrome-R66-10436.0.0-rc1/packages/Packages /tmp/tmpi25pAh.tmp That's this amd64-generic and this kahlee binhost package listing being copied temp temp files on disk (not clear to me where these temp files get read from). Picking an arbitrary package from BuildPackages that was build from source [ebuild N ] media-libs/fontconfig-2.11.1-r22:1.0::chromiumos to /build/kahlee/ USE="subpixel_rendering touchview -cros_host -doc -static-libs" ABI_X86="(64) -32 (-x32)" 0 KiB I see: It is present in the amd64 package listing but there's no tarball for it in the binhost tree. Are we failing to upload needed binary packages from amd64-generic paladin perhaps? It is present also in the kahlee package listing AND its tarball is present in the binhost tree. But perhaps this is being ignored because we are first searching in amd64 generic binhost and failing to use prebuilt if we are missing a tarball?
,
Feb 26 2018
,
Feb 26 2018
,
Feb 26 2018
,
Feb 26 2018
the Packages files get mirrored to /build/$BOARD/var/cache/edb/binhost/. they're copied to a /tmp file in case the download didn't go well. i haven't looked closer at the [ebuild] lines and why binpkgs weren't available. i do see a bunch of ARC++ packages in there that our amd64-generic builder wouldn't provide binpkgs for ... maybe we should add betty to the binhost mix.
,
Feb 26 2018
Note that the BinHostTestStage is supposed to fail in cases where we have configs that we don't have anyone uploading binpkgs for.
,
Feb 26 2018
pretty sure the BinhostTestStage only validates Chrome prebuilts. it doesn't really look at any other binpkg state.
,
Feb 26 2018
^ correct
,
Mar 10 2018
It sounds like this is fixed now that portage behaves properly?
,
Mar 10 2018
kahlee isn't always the slowest now, the current winner is cyan. http://shortn/_8Iy6kb2cM6 That doesn't necessarily mean that prebuilts are getting used correctly though.
,
Mar 10 2018
BuildPackages on kahlee down to ~27 minutes and mostly from prebuilts, so sure looks like we can mark this one as fixed. https://logs.chromium.org/v/?s=chromeos%2Fbb%2Fchromeos%2Fkahlee-paladin%2F1573%2F%2B%2Frecipes%2Fsteps%2FBuildPackages%2F0%2Fstdout |
|||||
►
Sign in to add a comment |
|||||
Comment 1 by dgarr...@chromium.org
, Feb 26 2018