[
https://issues.apache.org/jira/browse/CASSANDRA-14759?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16625894#comment-16625894
]
Benedict commented on CASSANDRA-14759:
--------------------------------------
bq. semantically, wouldn't it be better to leave full pending in pending and
filter natural transient out? It might be slightly confusing that natural
replica flips from transient to full: it doesn't seem as we have inherent
limitations that prevent us from that, or?
It might be, but unfortunately pending is used - almost exclusively - to
calculate how much we should boost our CL blockFor requirements by. The
sufficiency logic is the last bit of this codepath that could do with being
refactored, but until we do this, it is probably more dangerous to leave it in
the pending collection and risk boosting blockFor accidentally.
> Transient->Full movements mishandle consistency level upgrade
> -------------------------------------------------------------
>
> Key: CASSANDRA-14759
> URL: https://issues.apache.org/jira/browse/CASSANDRA-14759
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Reporter: Benedict
> Assignee: Benedict
> Priority: Major
> Labels: Availability, correctness, transient-replication
> Fix For: 4.0
>
>
> While we need treat a transitioning node as ‘full’ for writes, so that it can
> safely begin serving full data requests once it has finished, we cannot
> maintain it in the ‘pending’ collection else we will also increase our
> consistency requirements by a node that doesn’t exist.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
