From 8f1ce51e26dd24091f157af20aee9eafa7327c65 Mon Sep 17 00:00:00 2001 From: Or Balayla Date: Tue, 6 Oct 2026 16:31:37 +0300 Subject: [PATCH] ci: give the Coverity scan enough CPU and memory to run in parallel With a 4Gi limit cov-analyze revises from 4 workers to 1 due to memory constraints, so the scan alone takes 22-33 minutes. With 8 CPUs and 32Gi it runs 8 workers and takes about 4.5 minutes; the whole job takes about 5.5. Requests stay small (1 CPU, 4Gi), so the pod still schedules easily. Signed-off-by: Or Balayla --- .ci/pipeline/coverity_matrix.yaml | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/.ci/pipeline/coverity_matrix.yaml b/.ci/pipeline/coverity_matrix.yaml index b96754ce0..11c28b7e3 100644 --- a/.ci/pipeline/coverity_matrix.yaml +++ b/.ci/pipeline/coverity_matrix.yaml @@ -18,14 +18,16 @@ kubernetes: # runs_on_dockers entry below), same as the release pipeline - HPCINFRA-4329. runs_on_dockers: - # Same toolbox images and tags the release pipeline pins, with its - # requests/limits carried over unchanged. + # Same toolbox image and tag the release pipeline pins. Its 4 CPU/4Gi limits + # are raised: cov-analyze runs one worker per CPU and each needs its own + # memory, so under 4Gi it drops to 1 worker and the scan takes 22-33 + # minutes. 8 CPU/32Gi keeps 8 workers and takes under 5. - url: 'harbor.mellanox.com/hpcx/x86_64/coverity_2023:until_2028' name: 'coverity_scan' tag: 'base' arch: 'x86_64' - requests: '{memory: 1Gi, cpu: 1000m}' - limits: '{memory: 4Gi, cpu: 4000m}' + requests: '{memory: 4Gi, cpu: 1000m}' + limits: '{memory: 32Gi, cpu: 8000m}' steps: - name: Coverity scan