Merge pull request flux-framework#2736 from grondo/flux-module-reload

flux-module: add reload command

doc/man1/flux-module.adoc

@@ -41,6 +41,11 @@ Remove module 'name'.  The service that will unload the module is
 inferred from the name specified on the command line. If '-f, --force'
 is used, then do not error if module 'name' is not loaded.
 
+*reload* [--force] 'name' ['module-arguments' ...]::
+Reload module 'name'. This is equivalent to running 'flux module remove'
+followed by 'flux module load'. It is a fatal error if module 'name' is
+not loaded during removal unless the `-f, --force` option is specified.
+
 *list* ['service']::
 List modules loaded by 'service', or by flux-broker(1) if 'service' is unspecified.
 

src/cmd/flux-module.c

@@ -32,6 +32,7 @@ const int max_idle = 99;
 int cmd_list (optparse_t *p, int argc, char **argv);
 int cmd_remove (optparse_t *p, int argc, char **argv);
 int cmd_load (optparse_t *p, int argc, char **argv);
+int cmd_reload (optparse_t *p, int argc, char **argv);
 int cmd_info (optparse_t *p, int argc, char **argv);
 int cmd_stats (optparse_t *p, int argc, char **argv);
 int cmd_debug (optparse_t *p, int argc, char **argv);
@@ -101,13 +102,27 @@ static struct optparse_subcommand subcommands[] = {
       0,
       remove_opts,
     },
+    { "unload",
+      "[OPTIONS] module",
+      "Unload module",
+      cmd_remove,
+      OPTPARSE_SUBCMD_HIDDEN,
+      remove_opts,
+    },
     { "load",
       "[OPTIONS] module",
       "Load module",
       cmd_load,
       0,
       legacy_opts,
     },
+    { "reload",
+      "[OPTIONS] module",
+      "Reload module",
+      cmd_reload,
+      0,
+      remove_opts,
+    },
     { "info",
       "[OPTIONS] module",
       "Display module info",
@@ -140,7 +155,8 @@ int usage (optparse_t *p, struct optparse_option *o, const char *optarg)
     fprintf (stderr, "flux module subcommands:\n");
     s = subcommands;
     while (s->name) {
-        fprintf (stderr, "   %-15s %s\n", s->name, s->doc);
+        if (!(s->flags & OPTPARSE_SUBCMD_HIDDEN))
+            fprintf (stderr, "   %-15s %s\n", s->name, s->doc);
         s++;
     }
     exit (1);
@@ -265,12 +281,11 @@ char *getservice (const char *modname)
     return service;
 }
 
-int cmd_load (optparse_t *p, int argc, char **argv)
+static void module_load (flux_t *h, optparse_t *p, int argc, char **argv)
 {
     char *modname;
     char *modpath;
     int n;
-    flux_t *h;
     flux_future_t *f;
 
     if ((n = optparse_option_index (p)) == argc) {
@@ -291,8 +306,6 @@ int cmd_load (optparse_t *p, int argc, char **argv)
         n++;
     }
 
-    if (!(h = flux_open (NULL, 0)))
-        log_err_exit ("flux_open");
     if (!(f = flux_rpc_pack (h,
                              topic,
                              optparse_get_int (p, "rank", FLUX_NODEID_ANY),
@@ -314,28 +327,24 @@ int cmd_load (optparse_t *p, int argc, char **argv)
     json_decref (args);
     free (modpath);
     free (modname);
+}
+
+int cmd_load (optparse_t *p, int argc, char **argv)
+{
+    flux_t *h;
+    if (!(h = flux_open (NULL, 0)))
+        log_err_exit ("flux_open");
+    module_load (h, p, argc, argv);
     flux_close (h);
     return 0;
 }
 
-int cmd_remove (optparse_t *p, int argc, char **argv)
+static void module_remove (flux_t *h, const char *modname, optparse_t *p)
 {
-    char *modname;
-    flux_t *h;
     flux_future_t *f;
-    int n;
-
-    if ((n = optparse_option_index (p)) != argc - 1) {
-        optparse_print_usage (p);
-        exit (1);
-    }
-    modname = argv[n++];
-
     char *service = getservice (modname);
     char *topic = xasprintf ("%s.rmmod", service);
 
-    if (!(h = flux_open (NULL, 0)))
-        log_err_exit ("flux_open");
     if (!(f = flux_rpc_pack (h,
                              topic,
                              optparse_get_int (p, "rank", FLUX_NODEID_ANY),
@@ -351,10 +360,50 @@ int cmd_remove (optparse_t *p, int argc, char **argv)
     flux_future_destroy (f);
     free (topic);
     free (service);
+}
+
+int cmd_remove (optparse_t *p, int argc, char **argv)
+{
+    char *modname;
+    flux_t *h;
+    int n;
+
+    if ((n = optparse_option_index (p)) != argc - 1) {
+        optparse_print_usage (p);
+        exit (1);
+    }
+    modname = argv[n++];
+
+    if (!(h = flux_open (NULL, 0)))
+        log_err_exit ("flux_open");
+
+    module_remove (h, modname, p);
+
     flux_close (h);
     return (0);
 }
 
+int cmd_reload (optparse_t *p, int argc, char **argv)
+{
+    char *modname;
+    char *modpath;
+    flux_t *h;
+    int n;
+
+    if ((n = optparse_option_index (p)) == argc) {
+        optparse_print_usage (p);
+        exit (1);
+    }
+    parse_modarg (argv[n], &modname, &modpath);
+
+    if (!(h = flux_open (NULL, 0)))
+        log_err_exit ("flux_open");
+
+    module_remove (h, modname, p);
+    module_load (h, p, argc, argv);
+    return (0);
+}
+
 /* Translate an array of service names to a comma-delimited string.
  * If list is empty, NULL is returned.
  * If skip != NULL, skip over that name in the list (intended to be

t/t0003-module.t

@@ -18,6 +18,10 @@ test_expect_success 'module: load test module' '
 	flux module load \
 		${FLUX_BUILD_DIR}/t/module/.libs/parent.so
 '
+test_expect_success 'module: reload test module' '
+	flux module reload \
+		${FLUX_BUILD_DIR}/t/module/.libs/parent.so
+'
 
 test_expect_success 'module: lsmod shows test module' '
 	flux module list | grep parent
@@ -251,6 +255,7 @@ test_expect_success 'flux module -h lists subcommands' '
 	! flux module -h 2>module.help &&
 	grep -q list module.help &&
 	grep -q remove module.help &&
+	grep -q reload module.help &&
 	grep -q load module.help &&
 	grep -q info module.help &&
 	grep -q stats module.help &&

t/t0015-cron.t

@@ -274,11 +274,8 @@ test_expect_success 'flux-cron can set stop-on-failure' '
 '
 
 ##  Reload cron module with sync enabled
-test_expect_success 'flux module remove cron' '
-    flux module remove cron
-'
-test_expect_success 'module load with sync' '
-    flux module load cron sync=cron.sync sync_epsilon=0.025
+test_expect_success 'module reload with sync' '
+    flux module reload cron sync=cron.sync sync_epsilon=0.025
 '
 test_expect_success 'sync and sync_epsilon are set as expected' '
     flux cron sync | grep "cron\.sync.*epsilon=0.025"

t/t1003-kvs-stress.t

@@ -126,8 +126,7 @@ test_expect_success 'kvs: test that KVS_NO_MERGE works with kvs_commit()' '
 # transaction-merge option test
 test_expect_success 'kvs: transaction-merge disabling works' '
         THREADS=64 &&
-        flux module remove kvs &&
-        flux module load kvs transaction-merge=0 &&
+        flux module reload kvs transaction-merge=0 &&
         OUTPUT=`${FLUX_BUILD_DIR}/t/kvs/transactionmerge ${THREADS} $(basename ${SHARNESS_TEST_FILE})` &&
 	test "$OUTPUT" = "${THREADS}"
 '

t/t2200-job-ingest.t

@@ -152,8 +152,7 @@ test_expect_success 'submit request with empty payload fails with EPROTO(71)' '
 '
 
 test_expect_success 'job-ingest: test validator with version 1 enforced' '
-	flux exec -r all flux module remove job-ingest &&
-	flux exec -r all flux module load job-ingest \
+	flux exec -r all flux module reload job-ingest \
 		validator=${BINDINGS_VALIDATOR} validator-args="--require-version,1"
 '
 
@@ -162,8 +161,7 @@ test_expect_success 'job-ingest: v1 jobspecs accepted with v1 requirement' '
 '
 
 test_expect_success 'job-ingest: test non-python validator' '
-	flux exec -r all flux module remove job-ingest &&
-	flux exec -r all flux module load job-ingest \
+	flux exec -r all flux module reload job-ingest \
 		validator=${FAKE_VALIDATOR}
 '
 
@@ -172,8 +170,7 @@ test_expect_success 'job-ingest: submit succeeds with non-python validator' '
 '
 
 test_expect_success 'job-ingest: test python jsonschema validator' '
-	flux exec -r all flux module remove job-ingest &&
-	flux exec -r all flux module load job-ingest \
+	flux exec -r all flux module reload job-ingest \
 		validator=${JSONSCHEMA_VALIDATOR} validator-args=--schema,${SCHEMA}
 '
 
@@ -190,8 +187,7 @@ test_expect_success 'job-ingest: invalid jobs rejected by jsonschema validator'
 '
 
 test_expect_success 'job-ingest: validator unexpected exit is handled' '
-	flux exec -r all flux module remove job-ingest &&
-	flux exec -r all flux module load job-ingest \
+	flux exec -r all flux module reload job-ingest \
 		validator=${BAD_VALIDATOR} &&
 	test_must_fail flux job submit basic.json 2>badvalidator.out &&
 	grep "unexpectedly exited" badvalidator.out

t/t2202-job-manager.t

@@ -160,8 +160,7 @@ test_expect_success 'job-manager: that job is now the first job' '
 '
 
 test_expect_success 'job-manager: reload the job manager' '
-	flux module remove job-manager &&
-	flux module load job-manager
+	flux module reload job-manager
 '
 
 test_expect_success 'job-manager: queue was successfully reconstructed' '
@@ -229,8 +228,7 @@ test_expect_success 'job-manager: no jobs in the queue' '
 '
 
 test_expect_success 'job-manager: reload the job manager' '
-	flux module remove job-manager &&
-	flux module load job-manager
+	flux module reload job-manager
 '
 
 test_expect_success 'job-manager: still no jobs in the queue' '

t/t2203-job-manager-dummysched.t

@@ -95,8 +95,7 @@ test_expect_success 'job-manager: canceled job has exception, free events' '
 
 test_expect_success 'job-manager: reload sched-dummy --cores=4' '
 	flux dmesg -C &&
-	flux module remove sched-dummy &&
-	flux module load ${SCHED_DUMMY} --cores=4 &&
+	flux module reload ${SCHED_DUMMY} --cores=4 &&
 	flux dmesg | grep "hello_cb:" >hello.dmesg
 '
 

t/t2204-job-info.t

@@ -343,8 +343,7 @@ wait_inactive() {
 }
 
 test_expect_success 'reload the job-info module' '
-        flux exec -r all flux module remove job-info &&
-        flux exec -r all flux module load job-info &&
+        flux exec -r all flux module reload job-info &&
         wait_inactive
 '
 
@@ -435,8 +434,7 @@ test_expect_success 'flux job lists full path for job name if first argument not
 '
 
 test_expect_success 'reload the job-info module' '
-        flux exec -r all flux module remove job-info &&
-        flux exec -r all flux module load job-info
+        flux exec -r all flux module reload job-info
 '
 
 test_expect_success 'verify job names preserved across restart' '
@@ -471,8 +469,7 @@ test_expect_success 'flux job list outputs ntasks correctly (4 tasks)' '
 '
 
 test_expect_success 'reload the job-info module' '
-        flux exec -r all flux module remove job-info &&
-        flux exec -r all flux module load job-info
+        flux exec -r all flux module reload job-info
 '
 
 test_expect_success 'verify task count preserved across restart' '
@@ -525,8 +522,7 @@ test_expect_success 'flux job list outputs nnodes/ranks correctly (5 tasks, / 3
 '
 
 test_expect_success 'reload the job-info module' '
-        flux exec -r all flux module remove job-info &&
-        flux exec -r all flux module load job-info
+        flux exec -r all flux module reload job-info
 '
 
 test_expect_success 'verify nnodes preserved across restart' '

t/t2300-sched-simple.t

@@ -27,8 +27,7 @@ list_R() {
 }
 
 test_expect_success 'sched-simple: reload ingest module with lax validator' '
-        flux exec -r all flux module remove job-ingest &&
-	flux exec -r all flux module load job-ingest validator-args="--schema,${SCHEMA}" \
+	flux exec -r all flux module reload job-ingest validator-args="--schema,${SCHEMA}" \
          validator=${JSONSCHEMA_VALIDATOR}
 '
 test_expect_success 'sched-simple: generate jobspec for simple test job' '
@@ -103,8 +102,7 @@ test_expect_success 'sched-simple: cancel all jobs' '
 	test "$($query)" = "rank[0-1]/core[0-1]"
 '
 test_expect_success 'sched-simple: reload in best-fit mode' '
-	flux module remove sched-simple &&
-	flux module load sched-simple mode=best-fit
+	flux module reload sched-simple mode=best-fit
 '
 test_expect_success 'sched-simple: submit 5 more jobs' '
 	flux job submit basic.json >job6.id &&
@@ -161,8 +159,7 @@ test_expect_success 'sched-simple: check allocations for running jobs' '
 	test_cmp first-fit-allocs.expected first-fit-allocs.out
 '
 test_expect_success 'sched-simple: reload with outstanding allocations' '
-	flux module remove sched-simple &&
-	flux module load sched-simple &&
+	flux module reload sched-simple &&
 	flux dmesg | grep "hello: alloc rank0/core0" &&
 	test "$($query)" = ""
 '