[Tile-serving] [openstreetmap/mod_tile] Examine memory usage for potential leaks (Issue #445)

[Roland Bosa]

@hummeltech Oh, that merge was a surprise... I consider my branch still work in progress, and it had quite some junk in it. But now I see you just took the "good" parts and merged those. Thank you! I think the `jemalloc` choice is a step in the right direction!

I don't know if I've mentioned this already, but I managed ("hacked" - definitely don't merge that ;-) ) to get `renderd` to quit and dump a leak profile. But I'm having problems analyzing the profile in depth and won't have time for a while to work on that. But it looks though like there are no apparent leaks during the `renderd::mapnik` calls, it's just a lot of caching that sticks around.

So I started thinking of implementing some configurable recycling of the render threads, which might help to cut down on the overall memory usage; of course, at the cost of needing to reload the map style and fonts etc. Maybe with some tweak-able parameters to spread the cost of spinning up a render thread over N tiles, or M seconds, or both. I hope to get back to this when time allows...

-- 
Reply to this email directly or view it on GitHub:
https://github.com/openstreetmap/mod_tile/issues/445#issuecomment-2198623969
You are receiving this because you are subscribed to this thread.

Message ID: <openstreetmap/mod_tile/issues/445/2198623969 at github.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tile-serving/attachments/20240630/2e7d2119/attachment.htm>