thevar1able/ClickHouse
1
0
Fork 0
mirror of https://github.com/ClickHouse/ClickHouse.git synced 2024-11-22 07:31:57 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update website/blog/en/2022/clickhouse-v22.2-released.md

Browse Source 
Co-authored-by: Christoph Wurm <christoph@clickhouse.com>
This commit is contained in:
alexey-milovidov 2022-02-24 20:47:07 +03:00 committed by GitHub
parent 8c38c51d33
commit d0050f6ff4
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
website/blog/en/2022/clickhouse-v22.2-released.md
View File

@ -70,7 +70,7 @@ Another improvement is the option to specify TTL MOVE TO DISK/VOLUME **IF EXISTS
We split per-query and per-user memory limits into a pair of hard and soft limits. The settings `max_memory_usage` and `max_memory_usage_for_user` act as hard limits. When memory consumption is approaching the hard limit, an exception will be thrown. Two other settings: `max_guaranteed_memory_usage` and `max_guaranteed_memory_usage_for_user` act as soft limits. We split per-query and per-user memory limits into a pair of hard and soft limits. The settings `max_memory_usage` and `max_memory_usage_for_user` act as hard limits. When memory consumption is approaching the hard limit, an exception will be thrown. Two other settings: `max_guaranteed_memory_usage` and `max_guaranteed_memory_usage_for_user` act as soft limits.
A query will be allowed to use more memory than a soft limit if there is available memory. But if there will be memory shortage (with relative to per user hard limit or total per server memory consumption), we calculate the "overcommit ratio" - how more memory every query is consumed in relative to the soft limit, and we will kill the most overcommitted query to let other queries to run. A query will be allowed to use more memory than a soft limit if there is available memory. But if there will be memory shortage (relative to the per-user hard limit or total per-server memory consumption), we calculate the "overcommit ratio" - how much more memory every query is consuming relative to the soft limit - and we will kill the most overcommitted query to let other queries run.
In short, your query will not be limited to a few gigabytes of RAM if you have hundreds gigabytes available. In short, your query will not be limited to a few gigabytes of RAM if you have hundreds gigabytes available.