tcwg-cleanup-stale-containers: Rework to stop running containers

... instead of removing them.  This improvement makes cleanup job
to stop running containers after [default] 10 hours, and then
remove stopped containers after 10 days.

Change-Id: I1411a67788890b56cf6d08ee265beccdcdb7bd74

1 files changed, 34 insertions, 14 deletions

diff --git a/tcwg-cleanup-stale-containers.sh b/tcwg-cleanup-stale-containers.sh
index a3d27235..ca3f1a7f 100755
--- a/tcwg-cleanup-stale-containers.sh
+++ b/tcwg-cleanup-stale-containers.sh
@@ -2,22 +2,22 @@
 
 set -e
 
-docker_ps_opts=""
-hours="10"
-verbose=false
-cleanup_containers=false
-cleanup_volumes=false
+cleanup_running_hours="-10"
+cleanup_stopped_hours="-240"
 cleanup_images=false
+cleanup_volumes=false
+verbose=false
 
-OPTS="`getopt -l cleanup-containers:,cleanup-images:,cleanup-volumes:,docker_ps_opts:,hours:,verbose: -- "$@"`"
-while test $# -gt 0; do
+while [ $# -gt 0 ]; do
     case $1 in
-	--cleanup-containers) cleanup_containers="$2"; shift ;;
+	--cleanup-running-hours) cleanup_running_hours="$2"; shift ;;
+	--cleanup-stopped-hours) cleanup_stopped_hours="$2"; shift ;;
 	--cleanup-images) cleanup_images="$2"; shift ;;
 	--cleanup-volumes) cleanup_volumes="$2"; shift ;;
 	--docker_ps_opts) docker_ps_opts="$2"; shift ;;
 	--hours) hours="$2"; shift ;;
 	--verbose) verbose="$2"; shift ;;
+	*) echo "ERROR: Wrong option: $1"; exit 1 ;;
     esac
     shift
 done
@@ -26,10 +26,17 @@ if $verbose; then
     set -x
 fi
 
-if [ "$hours" -lt "0" ]; then
-    echo "ERROR: Refusing to delete containers that are $hours hours old"
-    exit 1
-fi
+do_cleanup_containers ()
+{
+    local hours="$1"
+    local docker_ps_opts="$2"
+    local action="$3"
+    local cleanup_containers=true
+
+    if [ "$hours" -lt "0" ]; then
+	hours="$((0-$hours))"
+	cleanup_containers=false
+    fi
 
 echo "Container report before:"
 docker ps $docker_ps_opts
@@ -56,8 +63,8 @@ if [ ${#rm_containers[@]} != 0 ]; then
     if $cleanup_containers; then
         for container in "${rm_containers[@]}"; do
             echo "Removing container $container"
-            docker rm -f -v $container | cat
-            echo "docker rm exit status: ${PIPESTATUS[0]}"
+            docker $action $container | cat
+            echo "docker $action exit status: ${PIPESTATUS[0]}"
         done
     else
         echo "DRY_RUN: NOT REMOVING CONTAINERS"
@@ -67,6 +74,19 @@ if [ ${#rm_containers[@]} != 0 ]; then
     docker ps $docker_ps_opts
 fi
 
+    exit $status
+}
+
+res="0"
+do_cleanup_containers $cleanup_running_hours "" "stop" &
+wait $! || res=$?
+status=$(($status+$res))
+
+res="0"
+do_cleanup_containers $cleanup_stopped_hours "-a" "rm -fv" &
+wait $! || res=$?
+status=$(($status+$res))
+
 if [ x"$(docker volume ls -q -f dangling=true)" != x"" ]; then
     echo "Removing dangling volumes"
     echo "Increasing exit code by 2 to indicate dangling volumes"