d9567272a5
This is a new config that will define the value to be used in parameters passed to different services as an amount of workers or threads to use. Most puppet modules default the amount of workers or threads to $::processorcount and this yields fairly high RAM usage on machines with a lot of cores. Since Packstack did not set this, we would use the default provided by the module. 8 cores with fairly minimal services yields >8GB RAM usage after tempest tests. If we set --service-workers to 2, we peak at around 6.5GB RAM for the same Packstack parameters. We will still default to $::processorcount for backwards compatibility but this new parameter can be used to lower the RAM usage where there are constraints (i.e, upstream gate) Additionally, instead of hardcoding a max_connections of 1024 which provides a ceiling of ~3GB RAM, we now instead default to a sane value depending on the amount of service workers (n*128). This will still yield the same 1024 max_connections for 8 cores but would set a max_connection of 256 when using 2 service workers for example. Change-Id: I2153ce522f227678d5318080dfb4ef171b36cbf7 |
||
|---|---|---|
| .. | ||
| conf.py | ||
| Makefile | ||
| packstack.rst | ||