[ https://issues.apache.org/jira/browse/CASSANALYTICS-194 ]
Saranya Krishnakumar deleted comment on CASSANALYTICS-194:
----------------------------------------------------
was (Author: saranya_k):
[https://github.com/apache/cassandra-analytics/commit/941b58fc5b02803995c8a75944b726e7bc92c0ed]
> Parse <replicas>/<transient> replication factor for witness-enabled keyspaces.
> ------------------------------------------------------------------------------
>
> Key: CASSANALYTICS-194
> URL: https://issues.apache.org/jira/browse/CASSANALYTICS-194
> Project: Apache Cassandra Analytics
> Issue Type: Improvement
> Components: Reader
> Reporter: Mansi Khara
> Assignee: Mansi Khara
> Priority: Normal
> Fix For: 0.5
>
> Time Spent: 4h 10m
> Remaining Estimate: 0h
>
> Adds support for the <replicas>/<transient> replication factor form, which
> witness replicas under mutation tracking reuse. A witness enabled keyspace
> declares 'datacenter1': '3/1', which means 3 replicas where 1 is a witness.
> Earlier CqlUtils.extractReplicationFactor() passed each value to
> Integer.parseInt, so a bulk read on any such keyspace failed at job setup
> with an uncaught NumberFormatException. We now track the transient count per
> datacenter and expose the full and transient counts through new accessors.
> getTotalReplicationFactor() still returns all replicas including witnesses,
> so nothing changes for untracked keyspaces. Validation matches what Cassandra
> does in locator.ReplicationFactor.validate. A new parseStrict factory reports
> a bad or empty replication map right away instead of dropping the datacenter
> and failing later with a confusing "DC not found in replication factor". The
> Kryo and CassandraRing JDK serializers are updated so the new field is not
> lost when it goes to Spark executors.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
