[gantry] return non-zero when pre-run post-run failed. skip updating …

…when pre-run failed.

README.md

@@ -43,7 +43,7 @@ You can configure the most behaviors of *Gantry* via environment variables.
 | GANTRY_LOG_LEVEL      | INFO | Control how many logs generated by *Gantry*. Valid values are `NONE`, `ERROR`, `WARN`, `INFO`, `DEBUG`. |
 | GANTRY_NODE_NAME      |      | Add node name to logs. If not set, *Gantry* will use the host name of the Docker Swarm's manager, which is read from either the Docker daemon socket of current node or `DOCKER_HOST`. |
 | GANTRY_POST_RUN_CMD   |      | Command(s) to `eval` after each updating iteration. For [example](examples/prune-and-watchtower), you can use this to remove unused containers, networks and images and update standalone docker containers. |
-| GANTRY_PRE_RUN_CMD    |      | Command(s) to `eval` before each updating iteration. For [example](examples/prune-and-watchtower), you can use this to remove unused containers, networks and images and update standalone docker containers. |
+| GANTRY_PRE_RUN_CMD    |      | Command(s) to `eval` before each updating iteration. For [example](examples/prune-and-watchtower), you can use this to remove unused containers, networks and images and update standalone docker containers. If you changed *Gantry* configurations in the pre-run command(s), the new value would apply to the following updating. If the last pre-run command failed, *Gantry* would skip updating services. |
 | GANTRY_SLEEP_SECONDS  | 0    | Interval between two updates. Set it to 0 to run *Gantry* once and then exit. When this is a non-zero value, after an updating, *Gantry* will sleep until the next scheduled update. The actual sleep time is this value minus time spent on updating services. |
 | TZ                    |      | Set timezone for time in logs. |
 

src/entrypoint.sh

@@ -121,6 +121,7 @@ gantry() {
   local ACCUMULATED_ERRORS=0
 
   eval_cmd "pre-run" "${PRE_RUN_CMD}"
+  ACCUMULATED_ERRORS=$((ACCUMULATED_ERRORS + $?))
 
   log INFO "Starting."
   gantry_initialize "${STACK}"
@@ -152,6 +153,7 @@ gantry() {
   ACCUMULATED_ERRORS=$((ACCUMULATED_ERRORS + $?))
 
   eval_cmd "post-run" "${POST_RUN_CMD}"
+  ACCUMULATED_ERRORS=$((ACCUMULATED_ERRORS + $?))
 
   local TIME_ELAPSED=
   TIME_ELAPSED=$(time_elapsed_since "${START_TIME}")

tests/gantry_common_options_spec.sh

@@ -155,7 +155,8 @@ Describe 'common-options'
     AfterEach "common_cleanup ${TEST_NAME} ${IMAGE_WITH_TAG} ${SERVICE_NAME}"
     It 'run_test'
       When run test_common_PRE_POST_RUN_CMD "${TEST_NAME}" "${SERVICE_NAME}"
-      The status should be success
+      # Updating should be successful, but post-run comamnd failed.
+      The status should be failure
       The stdout should satisfy display_output
       The stdout should satisfy spec_expect_no_message ".+"
       The stderr should satisfy display_output