Merge pull request #6598 from garlick/issue#6596

fix housekeeping reconfiguration problem
flux-framework · Feb 3, 2025 · a50142e · a50142e
2 parents ea5af0e + b748f94
commit a50142e
Show file tree

Hide file tree

Showing 2 changed files with 35 additions and 5 deletions.
diff --git a/src/modules/job-manager/housekeeping.c b/src/modules/job-manager/housekeeping.c
@@ -613,6 +613,7 @@ json_t *housekeeping_get_stats (struct housekeeping *hk)
     json_t *running;
     json_t *stats = NULL;
     struct allocation *a;
+    char *command = NULL;
 
     if (!(running = json_object ()))
         goto nomem;
@@ -626,11 +627,19 @@ json_t *housekeeping_get_stats (struct housekeeping *hk)
         }
         a = zlistx_next (hk->allocations);
     }
-    if (!(stats = json_pack ("{s:O}", "running", running)))
+    if (hk->cmd)
+        command = flux_cmd_stringify (hk->cmd);
+    if (!(stats = json_pack ("{s:O, s{s:f s:s}}",
+                             "running", running,
+                             "config",
+                               "release-after", hk->release_after,
+                               "command", command ? command : "")))
         goto nomem;
+    free (command);
     json_decref (running);
     return stats;
 nomem:
+    free (command);
     json_decref (running);
     errno = ENOMEM;
     return NULL;
@@ -742,7 +751,8 @@ static int housekeeping_parse_config (const flux_conf_t *conf,
     flux_error_t e;
     json_error_t jerror;
     json_t *cmdline = NULL;
-    const char *release_after = NULL;
+    const char *release_after_fsd = NULL;
+    double release_after = default_release_after;
     flux_cmd_t *cmd = NULL;
     const char *imp_path = NULL;
     char *imp_path_cpy = NULL;
@@ -764,7 +774,7 @@ static int housekeeping_parse_config (const flux_conf_t *conf,
                         0,
                         "{s?o s?s s?b !}",
                         "command", &cmdline,
-                        "release-after", &release_after,
+                        "release-after", &release_after_fsd,
                         "use-systemd-unit", &use_systemd_unit) < 0)
         return errprintf (error, "job-manager.housekeeping: %s", jerror.text);
 
@@ -778,8 +788,8 @@ static int housekeeping_parse_config (const flux_conf_t *conf,
     // let job-exec handle exec errors
     (void)flux_conf_unpack (conf, NULL, "{s?{s?s}}", "exec", "imp", &imp_path);
 
-    if (release_after) {
-        if (fsd_parse_duration (release_after, &hk->release_after) < 0)
+    if (release_after_fsd) {
+        if (fsd_parse_duration (release_after_fsd, &release_after) < 0)
             return errprintf (error,
                               "job-manager.housekeeping.release-after"
                               " FSD parse error");
@@ -813,6 +823,7 @@ static int housekeeping_parse_config (const flux_conf_t *conf,
     hk->cmd = cmd;
     free (hk->imp_path);
     hk->imp_path = imp_path_cpy;
+    hk->release_after = release_after;
     flux_log (hk->ctx->h,
               LOG_DEBUG,
               "housekeeping is %sconfigured%s",

diff --git a/t/t2226-housekeeping.t b/t/t2226-housekeeping.t
@@ -395,4 +395,23 @@ test_expect_success 'one node is allocated' '
 	test $(FLUX_RESOURCE_LIST_RPC=sched.resource-status \
 		flux resource list -s allocated -no {nnodes}) -eq 1
 '
+# issue #6596
+test_expect_success 'release-after is 0 from last test' '
+	flux module stats job-manager | \
+		jq -e .housekeeping.config >config1.json &&
+	jq -e ".\"release-after\" == 0" config1.json
+'
+test_expect_success 'reconfigure housekeeping with default release-after' '
+	flux config load <<-EOT
+	[job-manager.housekeeping]
+	command = [ "true" ]
+	# NOTE: release-after was "0" before, now it should be unset
+	EOT
+'
+test_expect_success 'release-after is -1 after reconfig' '
+	flux module stats job-manager | \
+		jq -e .housekeeping.config >config2.json &&
+	jq -e ".\"release-after\" == -1" config2.json
+'
+
 test_done