You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Blake Eggleston (JIRA)" <ji...@apache.org> on 2018/06/12 17:37:00 UTC
[jira] [Updated] (CASSANDRA-14517) Short read protection can cause
partial updates to be read
[ https://issues.apache.org/jira/browse/CASSANDRA-14517?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Blake Eggleston updated CASSANDRA-14517:
----------------------------------------
Fix Version/s: 4.0
> Short read protection can cause partial updates to be read
> ----------------------------------------------------------
>
> Key: CASSANDRA-14517
> URL: https://issues.apache.org/jira/browse/CASSANDRA-14517
> Project: Cassandra
> Issue Type: Bug
> Reporter: Blake Eggleston
> Priority: Major
> Fix For: 4.0
>
>
> If a read is performed in two parts due to short read protection, and the data being read is written to between reads, the coordinator will return a partial update. Specifically, this will occur if a single partition batch updates clustering values on both sides of the SRP break, or if a range tombstone is written that deletes data on both sides of the break. At the coordinator level, this breaks the expectation that updates to a partition are atomic, and that you can’t see partial updates.
>
> In some cases, read repair can make this partial update permanent. If a write hits a single node, but fails to reach the other replicas, part of it is returned via SRP and read repaired to the rest of the replicas, then the single node with then full write fails before repair or read repair, the partial write will become permanent.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org