CODEWITHSUNDEEP
cassandradatastaxcassandra-2.0datastax-enterprisecassandra-2.1
Naresh
Naresh

Reputation: 5397

how to reduce upgrade SSTable time in cassandra

I recently upgraded my Cassandra cluster from 2.1.12 to 3.9. And, it took 10-12 hours for each to upgrade the SStables total size of data was ~100GB on each node. Morever, compaction_throughput_mb_per_sec in my cluster was 16 as default value. What are the ways by means of which this time could have been reduced?

Upvotes: 1

Views: 366

Answers (1)

Chris Lohfink
Chris Lohfink

Reputation: 16400

Are you using SSDs? If you have extra IO bandwidth increasing compaction throughput (compaction_throughput_mb_per_sec) is the easiest. You can also increase the concurrent_compactors in your cassandra.yaml, but you will ultimately be limited by CPU and disk io.

Can also disable gossip/cql with nodetool on one node at a time to prevent it from impacting your application, then disable the compaction throughput all together. Just make sure to not have node removed from cluster for over 20 min (30 min is the default Hinted Handoff window).

Upvotes: 1

Related Questions