Ryan McGuire created CASSANDRA-6740:
---------------------------------------
Summary: Non-empty flush directory causes NPE on startup
Key: CASSANDRA-6740
URL: https://issues.apache.org/jira/browse/CASSANDRA-6740
Project: Cassandra
Issue Type: Bug
Reporter: Ryan McGuire
Priority: Minor
With the changes in CASSANDRA-6357, there is now the case where someone may
want to start a fresh cluster on a machine that previously hosted one, clean
out the directories they are familiar with (eg, data_file_directories,
commitlog_directories) but fail to clean out the flush_directory. If they don't
clean that out, they will see this in the logs:
{code}
INFO [main] 2014-02-19 12:20:13,530 ColumnFamilyStore.java:281 - Initializing
system.IndexInfo
INFO [main] 2014-02-19 12:20:13,534 ColumnFamilyStore.java:281 - Initializing
system.peers
INFO [main] 2014-02-19 12:20:13,539 ColumnFamilyStore.java:281 - Initializing
system.local
ERROR [main] 2014-02-19 12:20:13,703 CassandraDaemon.java:458 - Exception
encountered during startup
java.lang.NullPointerException: null
at
org.apache.cassandra.utils.ByteBufferUtil.string(ByteBufferUtil.java:167)
~[main/:na]
at
org.apache.cassandra.serializers.AbstractTextSerializer.deserialize(AbstractTextSerializer.java:39)
~[main/:na]
at
org.apache.cassandra.serializers.AbstractTextSerializer.deserialize(AbstractTextSerializer.java:26)
~[main/:na]
at
org.apache.cassandra.db.marshal.AbstractType.compose(AbstractType.java:66)
~[main/:na]
at
org.apache.cassandra.cql3.UntypedResultSet$Row.getString(UntypedResultSet.java:150)
~[main/:na]
at
org.apache.cassandra.config.CFMetaData.fromSchemaNoTriggers(CFMetaData.java:1761)
~[main/:na]
at
org.apache.cassandra.config.CFMetaData.fromSchema(CFMetaData.java:1883)
~[main/:na]
at
org.apache.cassandra.config.KSMetaData.deserializeColumnFamilies(KSMetaData.java:320)
~[main/:na]
at
org.apache.cassandra.config.KSMetaData.fromSchema(KSMetaData.java:301)
~[main/:na]
at
org.apache.cassandra.db.DefsTables.loadFromKeyspace(DefsTables.java:131)
~[main/:na]
at
org.apache.cassandra.config.DatabaseDescriptor.loadSchemas(DatabaseDescriptor.java:539)
~[main/:na]
at
org.apache.cassandra.service.CassandraDaemon.setup(CassandraDaemon.java:230)
[main/:na]
at
org.apache.cassandra.service.CassandraDaemon.activate(CassandraDaemon.java:441)
[main/:na]
at
org.apache.cassandra.service.CassandraDaemon.main(CassandraDaemon.java:530)
[main/:na]
{code}
I suggest a better warning message telling them it's due to the unclean flush
dir, instead of an NPE.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
