Problems with Rack Server and OpenStreetMap Nominatim. Any Advice?

Hi everyone,

I’m currently facing some issues with my rack server setup and an OpenStreetMap product. I’m relatively new to managing both hardware and software in this capacity and could really use some advice from those more experienced.

Here are some details about my setup:

  • Rack Server: Lenovo ThinkSystem SR650, Dual Intel Xeon Gold 5218, 64GB RAM, 2x 1TB NVMe SSD
  • OpenStreetMap Product: Nominatim (OpenStreetMap Geocoding Server)

The Issues:

  1. Performance Problems: Nominatim seems to run slowly on the server, especially during high-volume geocoding tasks. I’m not sure if this is due to a configuration issue or hardware limitations.
  2. Installation Challenges: I encountered several errors during the installation process of Nominatim on the server. While I managed to get it running, I’m worried that I might have missed some important steps or configurations.
  3. Data Sync Issues: There are difficulties in keeping the Nominatim database up-to-date with the latest OpenStreetMap data, leading to inconsistencies in geocoding results.

Steps I’ve Taken:

  • Verified that the server meets the minimum hardware requirements for Nominatim
  • Followed the installation guide provided by OpenStreetMap
  • Checked for any obvious configuration issues in the server and Nominatim settings

Despite these efforts, the issues persist. Has anyone else encountered similar problems with Nominatim? Are there specific settings or configurations I should be aware of? Any troubleshooting tips or resources you could recommend would be greatly appreciated.

Thanks in advance for your help!

I have never run Nominatim, so can’t comment on any specific issues with the software setup.

However, looking at the OpenStreetMap server hardware info page:

and specifically four mentioned Nominatim servers there:

all have considerably more RAM than your server. Even the oldest Xeon E5 v3 based platform (dulcy) still has 128GB, and the maximum there is 378GB (vhagar).

Having more RAM almost never hurts. Even if the database does not yet need all of it to cache data, the operating system will use free memory for caching files as well. Both speed up processing and keep CPUs busy doing real work instead of swapping data to disk if RAM is limited.

Having info about specific errors and how you solved them would help.

For high volume - order of magnitude info how many requests happen would also likely help


Also looking at the server data in Prometheus, does appear to show that both dulcy (128GB RAM), and stormfly (189GB RAM) have higher IO and memory pressure than the machines with more RAM:

And the Nominatim installation page also clearly recommends 128GB minimum for Planet:

so I guess if you can afford it, upgrading to something like 256GB RAM seems wise to alleviate RAM and disk bottlenecks if you run Planet.