[
https://issues.apache.org/jira/browse/CASSANDRA-20843?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18015567#comment-18015567
]
Ling Mao commented on CASSANDRA-20843:
--------------------------------------
Assign to me, I will make _*cached_rows*_ as guardrail and correct related
yaml/doc/comment/UT tonight
> Replica filtering guardrail needs a change in the yaml
> ------------------------------------------------------
>
> Key: CASSANDRA-20843
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20843
> Project: Apache Cassandra
> Issue Type: Bug
> Components: Local/Config
> Reporter: Paul Ayers
> Priority: Normal
> Labels: 5.0.5
> Fix For: 4.0.x, 4.1.x, 5.0.x, 6.x
>
>
> In the Cassandra 5 cassandra.yaml, we have the following guardrail section:
> {code:java}
> # Filtering and secondary index queries at read consistency levels above
> ONE/LOCAL_ONE use a
> # mechanism called replica filtering protection to ensure that results from
> stale replicas do
> # not violate consistency. (See CASSANDRA-8272 and CASSANDRA-15907 for more
> details.) This
> # mechanism materializes replica results by partition on-heap at the
> coordinator. The more possibly
> # stale results returned by the replicas, the more rows materialized during
> the query.
> replica_filtering_protection:
> # These thresholds exist to limit the damage severely out-of-date
> replicas can cause during these
> # queries. They limit the number of rows from all replicas individual
> index and filtering queries
> # can materialize on-heap to return correct results at the desired read
> consistency level.
> #
> # "cached_replica_rows_warn_threshold" is the per-query threshold at
> which a warning will be logged.
> # "cached_replica_rows_fail_threshold" is the per-query threshold at
> which the query will fail.
> #
> # These thresholds may also be adjusted at runtime using the
> StorageService mbean.
> #
> # If the failure threshold is breached, it is likely that either the
> current page/fetch size
> # is too large or one or more replicas is severely out-of-sync and in
> need of repair.
> cached_rows_warn_threshold: 2000
> cached_rows_fail_threshold: 32000
> {code}
> This is confusing for some as we had a customer attempt to configure
> cached_replica_rows_warn_threshold and cached_replica_rows_fail_threshold in
> addition to the actual parameters. We should update the wording so the
> description matches the parameters.
--
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
