From cb655d255466cb2f326ebb8546f37434b5bdb16d Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 6 Nov 2019 11:14:14 -0800 Subject: Revert "ci: Switch over to an autoscaling GKE cluster for builds." MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This reverts commit c9df92bf795af878c38538c85f781291c78ec513. It turns out that gitlab-runner uses kubernetes all wrong, spawning Pods and sshing into them to run the script instead of Jobs containing the script to run. This means that when anything goes wrong with the pod (autoscale, preemption, VM maintenance, cluster reconfiguration), the job fails and only sometimes gets handled as a runner system failure. Even worse, due to bugs in either the runner or k8s itself, some classes of timeout-related failure end up not being reported as failures, and the job will incorrectly report success! Disable using the "autoscale" cluster until we can do something else (docker-machine instead of k8s, or the custom third-party k8s-native runner). Reviewed-by: Michel Dänzer Acked-by: Daniel Stone --- .gitlab-ci/run-shader-db.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to '.gitlab-ci/run-shader-db.sh') diff --git a/.gitlab-ci/run-shader-db.sh b/.gitlab-ci/run-shader-db.sh index a62a91b5ea9..7580457d850 100755 --- a/.gitlab-ci/run-shader-db.sh +++ b/.gitlab-ci/run-shader-db.sh @@ -12,6 +12,6 @@ cd /usr/local/shader-db for driver in freedreno v3d; do env LD_PRELOAD=$LIBDIR/lib${driver}_noop_drm_shim.so \ - ./run -j 8 ./shaders \ + ./run -j 4 ./shaders \ > $ARTIFACTSDIR/${driver}-shader-db.txt done -- cgit v1.2.3