-
Notifications
You must be signed in to change notification settings - Fork 10
Description
and its kinda annoying to do by hand over and over again.
CardinalityExperiment.results.delete
INF [20101023-15:34:23.495] comm: Opening Zookeeper Connection to mesos-ec2.knowsql.org:2181 with timeout 30000
java.lang.RuntimeException: Server key schema {"type":"record","name":"ResultKey","namespace":"edu.berkeley.cs.scads.perf.scadr.cardinality","fields":[{"name":"clientConfig","type":{"type":"record","name":"LoadClient","fields":[{"name":"numServers","type":"int"},{"name":"numClients","type":"int"},{"name":"followingCardinality","type":"int"},{"name":"executorClass","type":"string"},{"name":"iterations","type":"int"},{"name":"threads","type":"int"},{"name":"runLengthMin","type":"int"}]}},{"name":"clusterAddress","type":"string"},{"name":"clientId","type":"int"},{"name":"iteration","type":"int"},{"name":"threadId","type":"int"}]} is not a prefix of client key schema {"type":"record","name":"ResultKey","namespace":"edu.berkeley.cs.scads.perf.scadr.cardinality","fields":[{"name":"clientConfi...
Activity
stephentu commentedon Oct 23, 2010
should schema mismatches then only emit warnings and not be fatal errors?
marmbrus commentedon Oct 23, 2010
One option would be to move delete out of namespace and into scads cluster
Another would be to make sure schema mistmatch doesn't fail in the constructor. I do think it should throw an exception when you try to get or put data though to ensure people don't corrupt the cluster.