podman/test/system/050-stop.bats
Ed Santiago d59391c047 system tests: fix race in stop test
In the unlock/timeout test, on slow systems, 'podman ps' could
catch the container before the just-backgrounded 'podman stop'
sends the signal. Wait for signal ack from container before
we inspect it.

Also: If I understand the test correctly, it wasn't actually
checking that 'ps' could grab the lock while the container
was exiting. Add a check.

Signed-off-by: Ed Santiago <santiago@redhat.com>
2021-07-29 11:57:51 -06:00

170 lines
6.1 KiB
Bash

#!/usr/bin/env bats
load helpers
# Very simple test
@test "podman stop - basic test" {
run_podman run -d $IMAGE sleep 60
cid="$output"
# Run 'stop'. Time how long it takes.
t0=$SECONDS
run_podman stop $cid
t1=$SECONDS
# Confirm that container is stopped. Podman-remote unfortunately
# cannot tell the difference between "stopped" and "exited", and
# spits them out interchangeably, so we need to recognize either.
run_podman inspect --format '{{.State.Status}} {{.State.ExitCode}}' $cid
is "$output" "\\(stopped\|exited\\) \+137" \
"Status and exit code of stopped container"
# The initial SIGTERM is ignored, so this operation should take
# exactly 10 seconds. Give it some leeway.
delta_t=$(( $t1 - $t0 ))
[ $delta_t -gt 8 ] ||\
die "podman stop: ran too quickly! ($delta_t seconds; expected >= 10)"
[ $delta_t -le 14 ] ||\
die "podman stop: took too long ($delta_t seconds; expected ~10)"
run_podman rm $cid
}
# #9051 : podman stop --all was not working with podman-remote
@test "podman stop --all" {
# Start three containers, create (without running) a fourth
run_podman run -d --name c1 $IMAGE sleep 20
run_podman run -d --name c2 $IMAGE sleep 40
run_podman run -d --name c3 $IMAGE sleep 60
run_podman create --name c4 $IMAGE sleep 80
# podman ps (without -a) should show the three running containers
run_podman ps --sort names --format '{{.Names}}--{{.Status}}'
is "${#lines[*]}" "3" "podman ps shows exactly three containers"
is "${lines[0]}" "c1--Up.*" "podman ps shows running container (1)"
is "${lines[1]}" "c2--Up.*" "podman ps shows running container (2)"
is "${lines[2]}" "c3--Up.*" "podman ps shows running container (3)"
# Stop -a
run_podman stop -a -t 1
# Now podman ps (without -a) should show nothing.
run_podman ps --format '{{.Names}}'
is "$output" "" "podman ps, after stop -a, shows no running containers"
# ...but with -a, containers are shown
run_podman ps -a --sort names --format '{{.Names}}--{{.Status}}'
is "${#lines[*]}" "4" "podman ps -a shows exactly four containers"
is "${lines[0]}" "c1--Exited.*" "ps -a, first stopped container"
is "${lines[1]}" "c2--Exited.*" "ps -a, second stopped container"
is "${lines[2]}" "c3--Exited.*" "ps -a, third stopped container"
is "${lines[3]}" "c4--Created.*" "ps -a, created container (unaffected)"
}
# #9051 : podman stop --ignore was not working with podman-remote
@test "podman stop --ignore" {
name=thiscontainerdoesnotexist
run_podman 125 stop $name
is "$output" \
"Error: no container with name or ID \"$name\" found: no such container" \
"podman stop nonexistent container"
run_podman stop --ignore $name
is "$output" "" "podman stop nonexistent container, with --ignore"
}
# Test fallback
# Regression test for #2472
@test "podman stop - can trap signal" {
# Because the --time and --timeout options can be wonky, try three
# different variations of this test.
for t_opt in '' '--time=5' '--timeout=5'; do
# Run a simple container that logs output on SIGTERM
run_podman run -d $IMAGE sh -c \
"trap 'echo Received SIGTERM, finishing; exit' SIGTERM; echo READY; while :; do sleep 1; done"
cid="$output"
wait_for_ready $cid
# Run 'stop' against it...
t0=$SECONDS
run_podman stop $t_opt $cid
t1=$SECONDS
# ...the container should trap the signal, log it, and exit.
run_podman logs $cid
is "$output" ".*READY.*Received SIGTERM, finishing" "podman stop $t_opt"
# Exit code should be 0, because container did its own exit
run_podman inspect --format '{{.State.ExitCode}}' $cid
is "$output" "0" "Exit code of stopped container"
# The 'stop' command should return almost instantaneously
delta_t=$(( $t1 - $t0 ))
[ $delta_t -le 2 ] ||\
die "podman stop: took too long ($delta_t seconds; expected <= 2)"
run_podman rm $cid
done
}
# Regression test for #8501
@test "podman stop - unlock while waiting for timeout" {
# Test that the container state transitions to "stopping" and that other
# commands can get the container's lock. To do that, run a container that
# ignores SIGTERM such that the Podman would wait 20 seconds for the stop
# to finish. This gives us enough time to try some commands and inspect
# the container's status.
run_podman run --name stopme -d $IMAGE sh -c \
"trap 'echo Received SIGTERM, ignoring' SIGTERM; echo READY; while :; do sleep 0.2; done"
wait_for_ready stopme
local t0=$SECONDS
# Stop the container, but do so in the background so we can inspect
# the container status while it's stopping. Use $PODMAN because we
# don't want the overhead and error checks of run_podman.
$PODMAN stop -t 20 stopme &
# Wait for container to acknowledge the signal. We can't use wait_for_output
# because that aborts if .State.Running != true
local timeout=5
while [[ $timeout -gt 0 ]]; do
run_podman logs stopme
if [[ "$output" =~ "Received SIGTERM, ignoring" ]]; then
break
fi
timeout=$((timeout - 1))
if [[ $timeout -eq 0 ]]; then
die "Timed out waiting for container to receive SIGERM"
fi
sleep 0.5
done
# Other commands can acquire the lock
run_podman ps -a
# The container state transitioned to "stopping"
run_podman inspect --format '{{.State.Status}}' stopme
is "$output" "stopping" "Status of container should be 'stopping'"
# Time check: make sure we were able to run 'ps' before the container
# exited. If this takes too long, it means ps had to wait for lock.
local delta_t=$(( $SECONDS - t0 ))
if [[ $delta_t -gt 5 ]]; then
die "Operations took too long ($delta_t seconds)"
fi
run_podman kill stopme
run_podman wait stopme
# Exit code should be 137 as it was killed
run_podman inspect --format '{{.State.ExitCode}}' stopme
is "$output" "137" "Exit code of killed container"
}
# vim: filetype=sh