You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
These changes will affect many users since most use default values.
Some suggestions:
cache-max-memory-size = 1048576000
Users with hot shards growing fast have had to increase this value. It should cause no harm raising it as it is only the maximum and can help avoid errors when hitting the max.
cache-snapshot-memory-size = 104857600
This is the in memory representation of cache, which can be different from the size on disk. Flushing the cache in larger chunks may yield better performance. There are indications from a few real world cases that this can help reduce I/O, but no hard numbers.
cache-snapshot-write-cold-duration = 10m
time allowed before cold cache gets flushed
compact-full-write-cold-duration = 4h
time allowed before a cold shard is compacted
Would be great to have performance testing figures with the old defaults and the new ones before decisions are made.
@jwilder could you add a note about how the demo config file works in terms of describing the current default values in comments? That will clear this issue up then #7504
This updates the config file to:
* Have all config options that do not change the default to be commented
out
* Removes old config options that are no longer available
* Updates config options to new defaults based on prior bugs and tuning
recommendations
Fixes#7320Fixes#7504
remove
compact-min-file-count
andmax-points-per-block
reduce
compact-full-write-cold-duration
andcache-snapshot-write-cold-duration
settingsincrease
cache-max-memory-size
?The text was updated successfully, but these errors were encountered: