Would it help if we have a seperate API to set a watch on a node, but the client sends the previous version of the node it saw as a parameter. When the server sets the watch, if the version of the node is higher than the version sent by the client, the watch will be triggered immediately. This should stop a client from doing a getChildren() (or getData()) each time it wants to set a watch and thus resulting in unconditional return of large amounts of data for each attempt to set a watch?
-Raghu ----- Original Message ---- From: Ted Dunning <ted.dunn...@gmail.com> To: zookeeper-user@hadoop.apache.org Sent: Friday, 8 May, 2009 14:14:22 Subject: Re: NodeChildrenChanged WatchedEvent If you really have thousands of nodes, you may be better off just collecting all the data into a single document rather than a directory of children. Zookeeper is fast, but making thousands of calls can eat up some time, especially if you are debugging a local client with a remote ZK server. On Fri, May 8, 2009 at 1:46 PM, Javier Vegas <jav...@beboinc.com> wrote: > My initial plan was to put thousands of children > under the same node, but it seems I will need to organize them on some > kind of hierarchical structure. >