From 3b50bb697121de96d03deffcd65ed5859de86558 Mon Sep 17 00:00:00 2001
From: Siri Hansen
Date: Fri, 7 Sep 2018 12:16:00 +0200
Subject: [supervisor] Add warning about race condition
ERL-724: "During a 'gentle' shutdown, supervisors unlink from their
children before sending shutdown signals to them. This can lead to a
race condition in supervision trees, when the timeout for gentle
shutdown of a parent supervisor expires and it kills a child
supervisor that has just unlinked from a child of its own, leaving the
child supervisor's own child still running after its supervisor is
killed."
This commit adds a warning about this in the documentation.
---
lib/stdlib/doc/src/supervisor.xml | 12 ++++++++++--
system/doc/design_principles/sup_princ.xml | 9 +++++++--
2 files changed, 17 insertions(+), 4 deletions(-)
diff --git a/lib/stdlib/doc/src/supervisor.xml b/lib/stdlib/doc/src/supervisor.xml
index 6d5065ca02..0e8075a578 100644
--- a/lib/stdlib/doc/src/supervisor.xml
+++ b/lib/stdlib/doc/src/supervisor.xml
@@ -208,8 +208,16 @@ child_spec() = #{id => child_id(), % mandatory
the child process is unconditionally terminated using
exit(Child,kill).
If the child process is another supervisor, the shutdown time
- is to be set to infinity to give the subtree ample
- time to shut down. It is also allowed to set it to infinity,
+ must be set to infinity to give the subtree ample
+ time to shut down.
+
+ Setting the shutdown time to anything other
+ than infinity for a child of type supervisor
+ can cause a race condition where the child in question
+ unlinks its own children, but fails to terminate them
+ before it is killed.
+
+ It is also allowed to set it to infinity,
if the child process is a worker.
Be careful when setting the shutdown time to
diff --git a/system/doc/design_principles/sup_princ.xml b/system/doc/design_principles/sup_princ.xml
index 06ca44a9f6..a2e0d95408 100644
--- a/system/doc/design_principles/sup_princ.xml
+++ b/system/doc/design_principles/sup_princ.xml
@@ -312,12 +312,17 @@ child_spec() = #{id => child_id(), % mandatory
signal back. If no exit signal is received within
the specified time, the child process is unconditionally
terminated using exit(Child, kill).
- - If the child process is another supervisor, it is to be
+
- If the child process is another supervisor, it must be
set to infinity to give the subtree enough time to
shut down. It is also allowed to set it to infinity,
- if the child process is a worker. See the warning below:
+ if the child process is a worker. See the warning below:
+ Setting the shutdown time to anything other
+ than infinity for a child of type supervisor
+ can cause a race condition where the child in question
+ unlinks its own children, but fails to terminate them
+ before it is killed.
Be careful when setting the shutdown time to
infinity when the child process is a worker. Because, in this
situation, the termination of the supervision tree depends on the
--
cgit v1.2.3