healthcheck_linux: avoid failing transient units

The main purpose of the transient services/timers is to trigger the
healthcheck execution in regular intervals, their own state should
not depend on the result of the healthchecks. This way there are no
failing systemd services unless there is actually a fatal error.

Signed-off-by: Patrick Wicki <patrick.wicki@siemens.com>

libpod/healthcheck_linux.go

@@ -47,7 +47,7 @@ func (c *Container) createTimer(interval string, isStartup bool) error {
 		cmd = append(cmd, "--log-level=debug", "--syslog")
 	}
 
-	cmd = append(cmd, "healthcheck", "run", c.ID())
+	cmd = append(cmd, "healthcheck", "run", "--ignore-result", c.ID())
 
 	conn, err := systemd.ConnectToDBUS()
 	if err != nil {

test/system/220-healthcheck.bats

@@ -101,12 +101,13 @@ Log[-1].Output   | \"Uh-oh on stdout!\\\nUh-oh on stderr!\\\n\"
     run -0 systemctl list-units
     cidmatch=$(grep "$cid" <<<"$output")
     echo "$cidmatch"
-    assert "$cidmatch" =~ " $cid-[0-9a-f]+\.timer  *.*/podman healthcheck run $cid" \
+    assert "$cidmatch" =~ " $cid-[0-9a-f]+\.timer  *.*/podman healthcheck run --ignore-result $cid" \
            "Healthcheck systemd unit exists"
 
     # Check that the right service option is applied so we don't hit the systemd restart limit.
     # Even though the code sets StartLimitIntervalSec the systemd command prints StartLimitInterval*U*Sec
-    run -0 systemctl show "$cid-*.service"
+    # Use show --all otherwise the glob might not match the already inactive transient unit.
+    run -0 systemctl show --all "$cid-*.service"
     assert "$output" =~ "StartLimitIntervalUSec=0" "The hc service has the right interval set"
 
     current_time=$(date --iso-8601=ns)