Refer to details of snapshot of raid problem.

diff --git a/doc/example.conf.in b/doc/example.conf.in
index 72ba35cec1f3ce5ae33aa09cadaa44455fe3d94d..ad8f2caa178503706a4f7e9dd9a3f76ab322a400 100644 (file)
--- a/doc/example.conf.in
+++ b/doc/example.conf.in
@@ -430,8 +430,12 @@ global {
     # "mirror" - The original RAID1 implementation provided by LVM2/DM.  It is
     #           characterized by a flexible log solution (core, disk, mirrored)
     #           and by the necessity to block I/O while reconfiguring in the
-    #           event of a failure.  Snapshots of this type of RAID1 can be
-    #           problematic.
+    #           event of a failure.
+    #
+    #           There is an inherent race in the dmeventd failure handling
+    #           logic with snapshots of devices using this type of RAID1 that
+    #           in the worst case could cause a deadlock.
+    #             Ref: https://bugzilla.redhat.com/show_bug.cgi?id=817130#c10
     #
     # "raid1"  - This implementation leverages MD's RAID1 personality through
     #           device-mapper.  It is characterized by a lack of log options.