tf-a-builder.yaml: resubmit LAVA jobs in case failure
Currently, at the TF LAVA lab there are two issues, not related to
user changes, impacting negatively the Open CI: Juno
infrastucture (corrupted SD cards, broken board, etc.)
[2] and FVP TFTF intermittent errors (TFTF execution hangs at some
particular test) [3]. The former (JUNO jobs) barely occurs and in most
cases the second submitted job passes correctly while the latter (FVP
TFTF jobs) issue is happening at least once in all top-level jobs,
causing false negatives reviews.
To cover both cases and logic does not get too complex, the approach
proposed consists of submitting the LAVA job, JUNO or FVP, in case of
failure at most tree times, reducing considerably the negative votes
given by the CI into user's gerrit review. The downside of this
approach is that the test phase may take longer (3 instead of 1 job
execution) but patch [4] proposes shorter and more realistic timeout
values, allowing LAVA to fail quicker due to [4] and re-trigger the
LAVA job again.
Finally, all this logic is a workaround while [3] is further investigated.
[1] https://tf.validation.linaro.org/
[2] https://linaro.atlassian.net/browse/STG-2489
[3] https://linaro.atlassian.net/browse/TFC-176
[4] https://review.trustedfirmware.org/c/ci/tf-a-ci-scripts/+/11759
Signed-off-by: Leonardo Sandoval <leonardo.sandoval@linaro.org>
Change-Id: Ib707100c605677c981ac1cd1cc08e6da146e8334
1 file changed