We have a CF with 11 secondary indexes (don't ask me why) and I noticed
restarting cassandra takes much longer time comparing to other clusters without
secondary indexes. In system.log I see 20 mins pause on building index.
this example shows a 12 mins gap.
INFO [SSTableBatchOpen:13] 2012-07-18 13:53:51,556 SSTableReader.java (line
153) Opening /data/cassandra/data/mobileks/MtsTrackingData-hc-5744 (1950145442
bytes)
INFO [SSTableBatchOpen:12] 2012-07-18 13:53:51,556 SSTableReader.java (line
153) Opening /data/cassandra/data/mobileks/MtsTrackingData-hc-5197
(12551211807 bytes)
INFO [OptionalTasks:1] 2012-07-18 14:05:27,648 SecondaryIndexManager.java
(line 183) Creating new index : ColumnDefinition{name=74696d657374616d70,
validator=org.apache.cassandra.db.marshal.DateType, index_type=KEYS,
index_name='MtsTrackingData_timestamp_idx'}
INFO [SSTableBatchOpen:2] 2012-07-18 14:05:27,673 SSTableReader.java (line
153) Opening
/data/cassandra/data/mobileks/MtsTrackingData.MtsTrackingData_timestamp_idx-hc-4354
(64493843 bytes)
INFO [SSTableBatchOpen:1] 2012-07-18 14:05:27,673 SSTableReader.java (line
153) Opening
/data/cassandra/data/mobileks/MtsTrackingData.MtsTrackingData_timestamp_idx-hc-4344
(258674041 bytes)
INFO [SSTableBatchOpen:5] 2012-07-18 14:05:27,673 SSTableReader.java (line
153) Opening
/data/cassandra/data/mobileks/MtsTrackingData.MtsTrackingData_timestamp_idx-hc-1826
(3397211685 bytes)
Is this by design? Why it has to be created during start up?
Feng Qu
