Andy Tolbert created CASSANDRA-20586:
----------------------------------------
Summary: Consider increasing default for
auto_repair.sstable_upper_threshold to a higher value
Key: CASSANDRA-20586
URL: https://issues.apache.org/jira/browse/CASSANDRA-20586
Project: Apache Cassandra
Issue Type: Improvement
Components: Consistency/Repair
Reporter: Andy Tolbert
Auto repair's {{sstable_upper_threshold}} default is currently defined as:
{noformat}
# # Threshold to skip repairing tables with too many SSTables. Defaults to
10,000 SSTables to avoid penalizing good
# # tables.
# sstable_upper_threshold: 10000
{noformat}
For clusters with denser nodes (multi-TB) and with LCS, 10k+ SSTables for a
given table is not unusual. It's probably best to have a higher threshold so
repairs don't completely stop working for a table, so maybe we should perhaps
set this to a default value that is higher and operators can decrease it
selectively?
A default between 30000-50000 would probably be more sensible.
--
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
