You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ozone.apache.org by "Ivan Andika (Jira)" <ji...@apache.org> on 2026/08/21 03:33:00 UTC
[jira] [Updated] (HDDS-13475) Follow RocksDB tuning guide on spinning disks
[ https://issues.apache.org/jira/browse/HDDS-13475?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ivan Andika updated HDDS-13475:
-------------------------------
Description:
See: [https://github.com/facebook/rocksdb/wiki/Tuning-RocksDB-on-Spinning-Disks] to tune RocksDB on spinning disks. This would be mostly catered to the container DB which is colocated on the DN data volume (which are usually on disk).
We might need to tune the DBProfile.DISK as follows
* Increase the block size: This might have memory overhead on the block cache
* Set DBOptions#setSkipStatsUpdateOnDbOpen to true
We need to have a performance benchmark to justify the tuning.
See https://sf-zhou.github.io/programming/rocksdb_direct_io_for_wal.html
was:
See: https://github.com/facebook/rocksdb/wiki/Tuning-RocksDB-on-Spinning-Disks to tune RocksDB on spinning disks. This would be mostly catered to the container DB which is colocated on the DN data volume (which are usually on disk).
We might need to tune the DBProfile.DISK as follows
* Increase the block size: This might have memory overhead on the block cache
* Set DBOptions#setSkipStatsUpdateOnDbOpen to true
We need to have a performance benchmark to justify the tuning.
> Follow RocksDB tuning guide on spinning disks
> ----------------------------------------------
>
> Key: HDDS-13475
> URL: https://issues.apache.org/jira/browse/HDDS-13475
> Project: Apache Ozone
> Issue Type: Improvement
> Reporter: Ivan Andika
> Priority: Major
>
> See: [https://github.com/facebook/rocksdb/wiki/Tuning-RocksDB-on-Spinning-Disks] to tune RocksDB on spinning disks. This would be mostly catered to the container DB which is colocated on the DN data volume (which are usually on disk).
> We might need to tune the DBProfile.DISK as follows
> * Increase the block size: This might have memory overhead on the block cache
> * Set DBOptions#setSkipStatsUpdateOnDbOpen to true
> We need to have a performance benchmark to justify the tuning.
> See https://sf-zhou.github.io/programming/rocksdb_direct_io_for_wal.html
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@ozone.apache.org
For additional commands, e-mail: issues-help@ozone.apache.org