-
Committer:
root
-
Date:
2012-03-07 08:17:57 UTC
-
Revision ID:
root@appscale-image0-20120307081757-6mc9gu3cv4m26ezq
Adding fault tolerance for appengine and using only a single watcher instead of a watcher for each role.
If a separate watcher is used for each role then if a client that created an ephemeral node being watched goes down then the connection to the zookeeper cluster is lost. (Is this a zk bug?)