It’s been intermittently very slow for me in JOSM over the last few weeks and was mentioned on #osm-gb
I really ought to use something like MapServer to provide this layer in JOSM locally, as I update my PostGIS copy of the England & Wales data when it’s published (1st Sunday of the month).
Using Josm with cadastral parcels turned on is torpedoing my aerial imagery - as soon as I turn on cadastrals the imagery stops loading and I get blank or heavily blurred tiles, when I disable cadastral parcels, it magically starts working again. Tested with Bing, Esri and Mapbox.
I’ve had similar issues with the Cadastral Parcels layer recently.
I think this is a really good example of how much a quietly developed service with little fanfare has such a huge positive impact that, when degraded, it causes problems for people.
I don’t believe there is already one, but if not, then a UK OSM service status page may be helpful. I am sure that not everyone would see or check it, but it might save some queries.
I’d like to thank the people behind the service who are quietly beavering away behind the scenes - I know I have found it immensely helpful.
Hello! some technical context for this:
The server providing this was hosted on AWS. Mythic beasts are generously donating hosting to osmuk and we have moved the tile server to a server hosted by them. I updated the DNS to point to the new server in early January.
As part of the move, some of the back end was changed (using postgres for the database instead of sqlite3) which should be much faster, but there is some kind of integration fault possibly between renderd and postgres that caused a resource leak, crashing the server quite regularly.
In the interests of stability, it was switched back to sqlite, but this is much slower as it is single threaded.
We’re seeing if there’s anything we can do to make it faster and more reliable again.
Follow Adge Cutler’s sage advice and bounce renderd daily (or twice if you prefer) and I suspect you’ll be fine
That said, I’m not convinced that there actually is a resource leak in the versions of everything that ships with recent Debian versions (specifically 12 and 13) - per the maintainer comment “even though the previously existing memory leak in PostgreSQL should have been resolved”.
What versions of renderd, postgres etc. are you running, and how many rendering threads?