diff --git a/docs/layouts/shortcodes/generated/cluster_configuration.html b/docs/layouts/shortcodes/generated/cluster_configuration.html index 1d1dbd87d9cf1d..ada54985eadf2f 100644 --- a/docs/layouts/shortcodes/generated/cluster_configuration.html +++ b/docs/layouts/shortcodes/generated/cluster_configuration.html @@ -62,6 +62,12 @@ Duration The shutdown timeout for cluster services like executors. + +
cluster.thread-dump.default-mode
+ "FULL" + String + Default granularity of the JobManager/TaskManager thread-dump REST endpoint when no explicit mode query parameter is supplied. Allowed values (case-insensitive): LITE (stack traces only, negligible JVM pause) and FULL (adds locked monitors and j.u.c. synchronizers, equivalent to jstack -l; pauses the JVM in a safepoint for a duration that scales with heap size and thread count -- seconds on large TaskManagers). The default is FULL to preserve historical behavior; operators of large clusters are strongly encouraged to switch to LITE to avoid heartbeat timeouts caused by long safepoint pauses. +
cluster.thread-dump.stacktrace-max-depth
50 diff --git a/docs/layouts/shortcodes/generated/expert_cluster_section.html b/docs/layouts/shortcodes/generated/expert_cluster_section.html index f32e45d44d0093..82b6b9364a403c 100644 --- a/docs/layouts/shortcodes/generated/expert_cluster_section.html +++ b/docs/layouts/shortcodes/generated/expert_cluster_section.html @@ -26,6 +26,12 @@ Boolean Whether processes should halt on fatal errors instead of performing a graceful shutdown. In some environments (e.g. Java 8 with the G1 garbage collector), a regular graceful shutdown can lead to a JVM deadlock. See FLINK-16510 for details. + +
cluster.thread-dump.default-mode
+ "FULL" + String + Default granularity of the JobManager/TaskManager thread-dump REST endpoint when no explicit mode query parameter is supplied. Allowed values (case-insensitive): LITE (stack traces only, negligible JVM pause) and FULL (adds locked monitors and j.u.c. synchronizers, equivalent to jstack -l; pauses the JVM in a safepoint for a duration that scales with heap size and thread count -- seconds on large TaskManagers). The default is FULL to preserve historical behavior; operators of large clusters are strongly encouraged to switch to LITE to avoid heartbeat timeouts caused by long safepoint pauses. +
cluster.thread-dump.stacktrace-max-depth
50 diff --git a/docs/layouts/shortcodes/generated/rest_v1_dispatcher.html b/docs/layouts/shortcodes/generated/rest_v1_dispatcher.html index 98c02d7a418fc5..26749b00f691b4 100644 --- a/docs/layouts/shortcodes/generated/rest_v1_dispatcher.html +++ b/docs/layouts/shortcodes/generated/rest_v1_dispatcher.html @@ -1383,6 +1383,16 @@ Returns the thread dump of the JobManager. + + Query parameters + + + + + +