...
This needs to be substantially refined for production.
The following section should address this in complete detail.
...
Performance Tuning
The EAS databases need to support 3 sorts of activities
- modest OLTP actvity (mad database)
- modest number of connections (maybe 20-30 simultaneuous?)
- most connections will be from single django web app with no connection pooling
- heavy weight overnight ETL activity (sfmaps, mad databases)
- heavy read processes from geoserver to produce web maps (sfmaps database)
- postgis is use (high IO)
The host VMs for the DB have 8GB of RAM.
We'll Until we have some real metrics, let's set aside 2GB for the OS and assume we have 6GB for the DB.
...
We'll need to do some monitoring in PROD to tune these values.
paramter | value | links (below) |
|---|---|---|
shared_buffers |
...
2GB |
...
1 |
...
,2 |
...
work_mem |
...
16MB |
...
2 |
...
maintenance_work_mem |
...
128MB |
...
2 |
...
wal_buffers |
...
1MB |
...
2 |
...
checkpoint_segments |
...
6 |
...
2 |
...
random_page_cost |
...
2 |
...
2 |
...
seq_page_cost |
...
1 |
...
2 |
...
effective_cache_size | 4GB |
...
1 |
...
max_connections |
...
100 |
...
default |
...
Do not forget to reload or restart for these configurations to take affect.
...
For any new users, we'll need to set the correct access privileges.
Until now, I have been using the postgres user everywhere (sorry!).
Performance Tuning
???
As I understand it, we do not want to use the default configurations.
We'll have to google a bit here I give it our best effort before PG Experts comes in.
Installation Errors
error:
...