diff --git a/systemd-mini.changes b/systemd-mini.changes
index 6a0a7640..25dbc3d6 100644
--- a/systemd-mini.changes
+++ b/systemd-mini.changes
@@ -1,3 +1,9 @@
+-------------------------------------------------------------------
+Thu Jul 29 13:12:48 UTC 2021 - Franck Bui <fbui@suse.com>
+
+- Avoid the error message when udev is updated due to udev being
+  already active when the sockets are started again (bsc#1188291)
+
 -------------------------------------------------------------------
 Mon Jul 26 10:54:10 UTC 2021 - Franck Bui <fbui@suse.com>
 
diff --git a/systemd-mini.spec b/systemd-mini.spec
index 6f2afebc..0f05b170 100644
--- a/systemd-mini.spec
+++ b/systemd-mini.spec
@@ -974,17 +974,21 @@ rm -f /etc/udev/rules.d/{20,55,65}-cdrom.rules
 
 %postun -n udev%{?mini}
 %regenerate_initrd_post
-systemctl daemon-reload || :
-# On package update, restarting the socket units will probably fail as
-# udevd is most likely running. Therefore systemctl will refuse to
-# start them again once stopped. It's not an issue since we are mostly
-# interested to make PID1 use the updated unit files once the socket
-# units wil be started again. And that will happen when systemd-udevd
-# itself will be restarted.
-if [ $1 -ge 1 ]; then
-	systemctl try-restart systemd-udevd-{control,kernel}.socket 2>/dev/null || :
-	systemctl try-restart systemd-udevd.service || :
-fi
+
+# The order of the units being restarted is important here because there's currently no
+# way to queue multiple jobs into a single transaction atomically. Therefore systemctl
+# will create 3 restart jobs that can be handled by PID1 separately and if the jobs for
+# the sockets are being handled first then starting them again will fail as the service
+# is still active hence the sockets held by udevd. However if the restart job for udevd
+# is handled first, there should be enough time to queue the socket jobs before the stop
+# job for udevd is processed. Hence PID1 will automatically sort the restart jobs
+# correctly by stopping the service then the sockets and then by starting the sockets and
+# the unit.
+#
+# Note that when systemd-udevd is restarted, there will always be a short time
+# frame where no socket will be listening to the events sent by the kernel, no
+# matter if the socket unit is restarted in first or not.
+%service_del_postun_with_restart systemd-udevd.service systemd-udevd-{control,kernel}.socket
 
 %posttrans -n udev%{?mini}
 %regenerate_initrd_posttrans
diff --git a/systemd.changes b/systemd.changes
index 6a0a7640..25dbc3d6 100644
--- a/systemd.changes
+++ b/systemd.changes
@@ -1,3 +1,9 @@
+-------------------------------------------------------------------
+Thu Jul 29 13:12:48 UTC 2021 - Franck Bui <fbui@suse.com>
+
+- Avoid the error message when udev is updated due to udev being
+  already active when the sockets are started again (bsc#1188291)
+
 -------------------------------------------------------------------
 Mon Jul 26 10:54:10 UTC 2021 - Franck Bui <fbui@suse.com>
 
diff --git a/systemd.spec b/systemd.spec
index f20d2bcf..f55e9088 100644
--- a/systemd.spec
+++ b/systemd.spec
@@ -972,17 +972,21 @@ rm -f /etc/udev/rules.d/{20,55,65}-cdrom.rules
 
 %postun -n udev%{?mini}
 %regenerate_initrd_post
-systemctl daemon-reload || :
-# On package update, restarting the socket units will probably fail as
-# udevd is most likely running. Therefore systemctl will refuse to
-# start them again once stopped. It's not an issue since we are mostly
-# interested to make PID1 use the updated unit files once the socket
-# units wil be started again. And that will happen when systemd-udevd
-# itself will be restarted.
-if [ $1 -ge 1 ]; then
-	systemctl try-restart systemd-udevd-{control,kernel}.socket 2>/dev/null || :
-	systemctl try-restart systemd-udevd.service || :
-fi
+
+# The order of the units being restarted is important here because there's currently no
+# way to queue multiple jobs into a single transaction atomically. Therefore systemctl
+# will create 3 restart jobs that can be handled by PID1 separately and if the jobs for
+# the sockets are being handled first then starting them again will fail as the service
+# is still active hence the sockets held by udevd. However if the restart job for udevd
+# is handled first, there should be enough time to queue the socket jobs before the stop
+# job for udevd is processed. Hence PID1 will automatically sort the restart jobs
+# correctly by stopping the service then the sockets and then by starting the sockets and
+# the unit.
+#
+# Note that when systemd-udevd is restarted, there will always be a short time
+# frame where no socket will be listening to the events sent by the kernel, no
+# matter if the socket unit is restarted in first or not.
+%service_del_postun_with_restart systemd-udevd.service systemd-udevd-{control,kernel}.socket
 
 %posttrans -n udev%{?mini}
 %regenerate_initrd_posttrans