[ https://issues.apache.org/jira/browse/CASSANDRA-19714?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Marcus Eriksson updated CASSANDRA-19714: ---------------------------------------- Change Category: Operability Complexity: Normal Component/s: Transactional Cluster Metadata Fix Version/s: 5.1-alpha1 Reviewers: Marcus Eriksson Assignee: Sam Tunnicliffe Status: Open (was: Triage Needed) https://github.com/krummas/cassandra/commit/0435a9dbc382a428864b4b329e127882d9c18419 > Use table-specific partitioners during Paxos repair > --------------------------------------------------- > > Key: CASSANDRA-19714 > URL: https://issues.apache.org/jira/browse/CASSANDRA-19714 > Project: Cassandra > Issue Type: Improvement > Components: Transactional Cluster Metadata > Reporter: Sam Tunnicliffe > Assignee: Sam Tunnicliffe > Priority: Normal > Fix For: 5.1-alpha1 > > > Partition keys in the \{{system.paxos}} table are derived from the key > involved in the paxos transaction. Initially, it was safe to assume that the > paxos table itself used the same partitioner as the tables in the > transactions as all distributed keyspaces and tables were configured with the > global partitioner. This is no longer true as the > \{{system_cluster_metadata.distributed_metadata_log}} has its own custom > partitioner. > Likewise, \{{PaxosRepairHistory}} and the \{{system.paxos_repair_history}} > table which makes that history durable map token ranges in the transacted > tables to ballots. Prior to CASSANDRA-19482 it was safe to assume that these > ranges contained tokens from the global partitioner but as this is no longer > the case, we must use the specific partitioner for the table in question when > working with ranges during paxos repair. -- This message was sent by Atlassian Jira (v8.20.10#820010) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org