[
https://issues.apache.org/jira/browse/CASSANDRA-1391?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13098252#comment-13098252
]
Pavel Yaskevich commented on CASSANDRA-1391:
--------------------------------------------
bq. Why do we need to do two code paths here? My preference would be to model
this on our Row conflict resolution: for rows, we have a single code path where
distinct columns are simply merged, and for conflicting columns we pick a
winner based on user-provided timestamp and, if necessary, value contents. So
the result is guaranteed to be the same on all replicas no matter what order
updates were received in.
Migration merging is more complex process comparing to row merging which is
pretty straight-forward, current approach easily handles all possible conflits
without any tie-breakers or coordinators because it simply detects what
modifications where made by each of the migrations starting from merging one,
combines them (modifications) together in isolated schema and updates
Schema.instance _so the resulting schema is guaranteed to be the same on all
replicas no matter what order migrations were received in_.
bq. Seems to me this would make the "isolation" complexity go away.
I still think that this is the simplest solution of all proposed because actual
modifications are: KSMetaData/CFMetaData.diff(...) methods to detect modified
fields, one flag member Migration.isolated to indicate that migration is
running in the isolated mode and one method to update system Schema.instance
with resulting Schema after all migrations where applied.
> Allow Concurrent Schema Migrations
> ----------------------------------
>
> Key: CASSANDRA-1391
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1391
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Stu Hood
> Assignee: Pavel Yaskevich
> Fix For: 1.0
>
> Attachments: CASSANDRA-1391.patch
>
>
> CASSANDRA-1292 fixed multiple migrations started from the same node to
> properly queue themselves, but it is still possible for migrations initiated
> on different nodes to conflict and leave the cluster in a bad state. Since
> the system_add/drop/rename methods are accessible directly from the client
> API, they should be completely safe for concurrent use.
> It should be possible to allow for most types of concurrent migrations by
> converting the UUID schema ID into a VersionVectorClock (as provided by
> CASSANDRA-580).
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
