[
https://issues.apache.org/jira/browse/CASSANDRA-11715?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15387242#comment-15387242
]
Erick Ramirez commented on CASSANDRA-11715:
-------------------------------------------
*FOR DOCS UPDATE*
New property in {{cassandra.yaml}} :
{noformat}
# By default, Cassandra logs GC Pauses greater than 200 ms at INFO level
# This threshold can be adjusted to minimize logging if necessary
# gc_log_threshold_in_ms: 200
{noformat}
Set to same value as G1 GC target pause in {{cassandra-env.sh}} (default is
500):
{noformat}
# Main G1GC tunable: lowering the pause target will lower throughput and
vise versa.
# 200ms is the JVM default and lowest viable setting
# 1000ms increases throughput. Keep it smaller than the timeouts in
cassandra.yaml.
JVM_OPTS="$JVM_OPTS -XX:MaxGCPauseMillis=500"
{noformat}
or {{jvm.options}}:
{noformat}
## Main G1GC tunable: lowering the pause target will lower throughput and vise
versa.
## 200ms is the JVM default and lowest viable setting
## 1000ms increases throughput. Keep it smaller than the timeouts in
cassandra.yaml.
#-XX:MaxGCPauseMillis=500
{noformat}
> Make GCInspector's MIN_LOG_DURATION configurable
> ------------------------------------------------
>
> Key: CASSANDRA-11715
> URL: https://issues.apache.org/jira/browse/CASSANDRA-11715
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Brandon Williams
> Assignee: Jeff Jirsa
> Priority: Minor
> Labels: docs-impacting, lhf
> Fix For: 2.2.8, 3.0.9, 3.8
>
>
> It's common for people to run C* with the G1 collector on appropriately-sized
> heaps. Quite often, the target pause time is set to 500ms, but GCI fires on
> anything over 200ms. We can already control the warn threshold, but these
> are acceptable GCs for the configuration and create noise at the INFO log
> level.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
