I upgraded to 1.0.6 to 1.1.0, and I noticed the effective ownership report
changed.
I have a 3-node cluster, with evenly divided tokens and RF=2. The node tool
report on 1.0.6 was:
33.33% 0
33.33% 56713727820156410577229101238628035243
33.33% 113427455640312821154458202477256070485
Under 1.1.0 it is
66.67% 0
66.67% 56713727820156410577229101238628035243
66.67% 113427455640312821154458202477256070485
Does the updated reporting in 1.1.0 include the replicated data and before it
didn't?
As long as I'm posting, I'll report other anomalies that I witnessed but
overcame.
I eventually gave up trying to do a live rolling update because it complained
certain column families didn't exist. Like this: This doesn't appear to be
any CF I am using.
ERROR [ReadStage:69680] 2012-05-02 01:54:14,995 AbstractCassandraDaemon.java
(line 133) Fatal exception
in thread Thread[ReadStage:69680,5,main]
java.lang.RuntimeException: java.lang.IllegalArgumentException: Unknown CF 1851
at
org.apache.cassandra.service.StorageProxy$DroppableRunnable.run(StorageProxy.java:1238)
at
java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1110)
at
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:603)
at java.lang.Thread.run(Thread.java:722)
Caused by: java.lang.IllegalArgumentException: Unknown CF 1851
at org.apache.cassandra.db.Table.getColumnFamilyStore(Table.java:167)
at org.apache.cassandra.db.Table.getColumnFamilyStore(Table.java:160)
at org.apache.cassandra.db.Table.getRow(Table.java:374)
at
org.apache.cassandra.db.SliceByNamesReadCommand.getRow(SliceByNamesReadCommand.java:58)
at
org.apache.cassandra.service.StorageProxy$LocalReadRunnable.runMayThrow(StorageProxy.java:766
)
at
org.apache.cassandra.service.StorageProxy$DroppableRunnable.run(StorageProxy.java:1234)
... 3 more
I tried repeatedly after various nodetool incantations but never succeeded.
Although I know it -can- work. It worked at least once, but I had to rollback
to 1.0.6 though, because I found 1.1.0 was using the new SnappyCompression.
But my target OS is Solaris, and there's only 3 included libraries in the jar
for SnappyCompression - Solaris not included. So I had to update all my column
family creation code to explicitly set compression to the previous JavaDeflate
default. That new default was annoying.
Ron
