Hi, I'm Patrick from the ZooKeeper team. Recently we've (the ZK dev
community) been working more closely with the HBase dev team. In
particular to ensure that ZK-HBase interaction is the best it can be and
improve things where it's not. We are also looking at how HBase might
take more advantage of ZK in future. Please see the following for some
running commentary (esp see the "use cases" link):
http://wiki.apache.org/hadoop/ZooKeeper/HBaseAndZooKeeper
One of the issues we identified early on is that HBase users don't
always know what to expect from ZooKeeper wrt scaling. As a result of
that discussion I created the following document which details ZK
scalability with varying CPUs available (1,2,4 cores) while varying the
client load and ZK configuration. See the summary section near the top
for some graphs that give a good overview (note: the load used in this
test is at least a couple orders of magnitude great than that applied by
HBase currently):
http://bit.ly/4ekN8G
If you are operating a ZK cluster, or would just like to know more about
ZK operations (esp monitoring) you might also want to review the
ZooKeeper operations manual:
http://hadoop.apache.org/zookeeper/docs/current/zookeeperAdmin.html
We also have a running wiki page to help with troubleshooting ZK issues.
This is based on problems users have had in the past, it also gives some
ideas on how to troubleshoot issues if you do experience problems:
http://wiki.apache.org/hadoop/ZooKeeper/Troubleshooting
I hope you find these useful. If you'd like more detail on ZooKeeper I
encourage you to subscribe to the ZK user list:
http://hadoop.apache.org/zookeeper/mailing_lists.html#Users
You can also follow me on http://twitter.com/phunt where I occasionally
post issues of interest to the ZK/Hadoop community.
Regards,
Patrick