fix remote run/start flake

Fix the flake reported in #9597 with a workaround to at least stop
wasting energy until the root cause has been found and fixed.

It seems that a remote run returns before the container has transitioned
into the `exited` state which ultimately breaks a subsequent remote
start with attach.

Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
This commit is contained in:
Valentin Rothberg 2021-12-06 15:22:55 +01:00 • committed by Ed Santiago
parent 9bf0271e6d
commit 921e9ecbb3

View file

@ -91,6 +91,10 @@ ${cid[0]} d" "Sequential output from logs"
function _log_test_restarted() {
run_podman run --log-driver=$1 --name logtest $IMAGE sh -c 'start=0; if test -s log; then start=`tail -n 1 log`; fi; seq `expr $start + 1` `expr $start + 10` | tee -a log'
# FIXME: #9597
# run/start is flaking for remote so let's wait for the container condition
# to stop wasting energy until the root cause gets fixed.
run_podman container wait --condition=exited logtest
run_podman start -a logtest
logfile=$(mktemp -p ${PODMAN_TMPDIR} logfileXXXXXXXX)
$PODMAN $_PODMAN_TEST_OPTS logs -f logtest > $logfile