Project status

[kszafran]

Hello @gdey and @ARolek!

In my company we're currently investigating Tegola to serve vector tiles from our PostGIS database. I wonder if you could update me on the finding from this "experiment". Did you find for example performance improvements by using PostGIS itself to simplify geometries and build MVT tiles? Are there plans on merging these changes into the main Tegola repository?

Thanks!

[ARolek]

@kszafran great to hear! This experiment was a success and it's planned to make its way back into the main tegola repo. Yesterday I cut a release (v0.11.0) and I'm looking at migrating this code into v0.12.0. Any interest in helping with the migration or testing?

[kszafran]

Hello. We'll at least test this version (tegola-postgis). So far the performance we got out of tegola was abysmal and we're experimenting trying to understand why and how to improve it. One challenge is that our data is constantly changing. I was thinking that we could for example cache lower zoom levels, where the changes wouldn't be that significant, and not cache higher zoom levels at all. That's what we still need to experiment with.

[ARolek]

@kszafran there are a ton of strategies for performance tuning but it's largely situational. The first place to look is at geometry indexes. Do you have those in place? One way to check if the issue is on tegola's side or at the database is to use run tegola with: TEGOLA_SQL_DEBUG=EXECUTE_SQL tegola --config=/path/to/conf.toml

This will output the SQL that tegola is using. You can then run the query directly at your database and see how long the query is taking.

I'm happy to run through other performance suggestions too. If you're interested in talking real time you can join our Slack: http://slack.go-spatial.org/

[gdey]

FYI: here is the pull request to merge this experiment into the mainline.

go-spatial/tegola#687