redesign how reloadnodes is implemented.
authorRonnie Sahlberg <ronniesahlberg@gmail.com>
Tue, 2 Dec 2008 02:26:30 +0000 (13:26 +1100)
committerRonnie Sahlberg <ronniesahlberg@gmail.com>
Tue, 2 Dec 2008 02:26:30 +0000 (13:26 +1100)
commitd1057ed6de7de9f2a64d8fa012c52647e89b515b
tree4a5c6a8d3e2545fd6557762b2e2833c2d2f833aa
parente577a276900854622f4e9da9d1ccd7b484d0d1ec
redesign how reloadnodes is implemented.

modify the transport methods to allow to restart individual connections
and set up destructors properly.

only tear down/set-up tcp connections to nodes removed from the cluster
or nodes added to the cluster.
Leave tcp connections to unchanged nodes connected.

make "ctdb reloadnodes" explicitely cause a recovery of the cluster once
the files have been realoaded
include/ctdb_private.h
server/ctdb_recover.c
tcp/tcp_connect.c
tcp/tcp_init.c
tools/ctdb.c