hcq/MaxScale
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix to MXS-436: https://mariadb.atlassian.net/browse/MXS-436

Browse Source 
Added autodetection of processor cores and used it as the default if an
invalid value is used for threads.
This commit is contained in:
Markus Makela
2015-10-31 07:36:22 +02:00
parent 205406f65c
commit bc148a4301
1 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
Getting-Started/Configuration-Guide.md
View File

@ -50,11 +50,13 @@ Please see the section about [Protocol Modules](#protocol-modules) for more deta
### Global Settings
The global settings, in a section named `[MaxScale]`, allow various parameters that affect MaxScale as a whole to be tuned. Currently the only setting that is supported is the number of threads to use to handle the network traffic. MaxScale will also accept the section name of `[gateway]` for global settings. This is for backward compatibility with versions prior to the naming of MaxScale.
The global settings, in a section named `[MaxScale]`, allow various parameters that affect MaxScale as a whole to be tuned.
#### `threads`
To control the number of threads that poll for network traffic set the parameter threads to a number. It is recommended that you start with a single thread and add more as you find the performance is not satisfactory. MaxScale is implemented to be very thread efficient, so a small number of threads is usually adequate to support reasonably heavy workloads. Adding more threads may not improve performance and can consume resources needlessly.
This parameter controls the number of threads that poll for network traffic. MaxScale will autodetect the number of processors of the system unless number of threads is manually configured. It is recommended that you let MaxScale detect how many cores the system has and leave this parameter undefined. The number of used cores will be logged into the message logs and if you are not satisfied with the autodetected value, you can manually configure it.
If you want to fine-tune the number of threads, start with a single thread and add more as you find the performance is not satisfactory. MaxScale is implemented to be very thread efficient, so a small number of threads is usually adequate to support reasonably heavy workloads. Adding more threads may not improve performance and can consume resources needlessly. Increasing the amount of worker threads beyond the number of processor cores is not recommended.
```
# Valid options are: