Some tiles ONLY fail to load in Safari/WebKit

[hdlnii]

Main tile server (tile.opentopomap.org) intermittently fails to load specific tiles in Safari (Mac) and WebKit-based apps (iOS) — "couldn't establish a secure connection" / TLS reset.

Curiously, the same tiles load perfectly fine in Chrome and Firefox on the same Mac, same network.
Backup server (backup.opentopomap.org) loads the failing tiles (mostly) without issue.

But on iOS there is no alternative than WebKit? So the main server is 100% unusable at the moment on iOS.

Checked SSL Labs, found no cert chain issues reported. Just thought it might be a good idea to bring this to your attention.
Related to the server load/resource situation in #382 maybe?

To reproduce

• load any mid-zoom tile in Safari a few times, some will fail while others succeed. Same URL works via backup server or in Chrome.

What I could reproduce consistently in my environment

macOS: Chrome/Safari/Firefox — on a same machine trying to load the same tile , and only Safari fails

iOS: the same tile not loading in any way in Safari acorss 4 tabs

iOS: some particular tiles not loading in a WebKit-based app I use everyday
(This is actually the place where I initially encountered and noticed this issue)

[krusty82]

A few questions:
Is it always the same tiles that are not loading?
Is it worse or better for different times of the day? (Could be an overloaded server)
backup.opentopomap.org and tile opentopomap.org are different servers.
Could you try tile.top-o-map.org as tile-source? It's the same server, but managed by cloudserve.
Is it happening with all zoomlevels?

[hdlnii]

Is it always the same tiles that are not loading?
Is it worse or better for different times of the day? (Could be an overloaded server)

What I do know so far is that this particular tile, right here: https://tile.opentopomap.org/6/46/25.png, which is the same one in my OP — I just can't get it open and loaded in Safari from yesterday's evening until today's afternoon, on both my Mac and my iOS. So this one is pretty consistent, at least across two days at different times.

Do you have a WebKit environment close to your hands that you can test this? Can you open this tile in it?
Because so far, I only have feedback and observations in my own environment, with the same networks, and etc. Frankly, I do not know for sure if this is a server-issue or a my-environemtn-issue, or that this is an issue only present under WebKit environments.

Is it happening with all zoomlevels?

Yes. I see irregular "Tile unavailable" white tiles appearng across different zoom levels, just like in:
https://github.com/user-attachments/assets/04b30af9-a7f9-4245-9996-71addb953c5e
and
https://github.com/user-attachments/assets/58adeace-ca8f-424a-9a68-a325dd90b016

backup.opentopomap.org and tile opentopomap.org are different servers.

I did learn about that. But I wonder why backup has a differenet carto style than the main one? and the zoom level is around 13~14 in my city in the backup, as opposed to the 17 (known) in main? Is backup a little more outdated, maybe it's an older rendering batch?
In the app I use everyday on iOS, backup loads mostly fine, but main server is unusable.

Could you try tile.top-o-map.org as tile-source? It's the same server, but managed by cloudserve.

I can't seem to get it working. Is this tile URL written correctly?

[krusty82]

I can't seem to get it working. Is this tile URL written correctly?

Sorry, should be tile.top-o-map.de

[hdlnii]

just tested https://tile.top-o-map.de/{z}/{x}/{y}.png as a tile source,, no luck still

[krusty82]

I changed some settings on top-o-map.de. could you please check again?
Also, I don't really know what the problem is, and I don't have a webkit environment (yet ;) ).

[krusty82]

When was the first time you noticed the error?

[hdlnii]

Tested https://tile.top-o-map.de/{z}/{x}/{y}.png in this mapping app I use every day, looking at different parts of the globe at different zooming levels. And it's such good news for me (!!!):

the specific tiles that are consistently unable to be loaded under https://tile.opentopomap.org/{z}/{x}/{y}.png (LEFT), are all successfully being loaded under top-o-map.de(RIGHT).

Thank you so so much! <3

May i ask,
Is this the same map as tile.opentopomap.org ?
What did you change? Curious to know whay might be causing this Issue.

As to the first time I noticed the error, I don't recall exact date, might have been just the recent couple of weeks, but this Issue might have existed before I noticed it.

---

I just checked out the website at https://top-o-map.de/, I think it is beautiful !
So, is this Thomas?

Though I am a little confused as to the relationship between Top-O-Map and OpenTopoMap.
Is it the same map? Same server as you said, so: same data and same set of rendered tiles, but just 2 seperate domains?
Is it a successor to OTM (since OTM has had challenges maintaining freshness) ?
Or is it a seperate brand/service on its own (as it seems to be on the website) ?

Thanks again,
The issue is basically solved in my practice, as I can just fetch Top-O-Map instead, under Webkit, whenever OTM fails. But unsure if OTM will ever fix this as it might still affect more people.
The rest is just curiosity as I am intrigued.

[arachsys]

You can demonstrate the problem here with

curl --http2 -v https://c.tile.opentopomap.org/8/125/81.png -o /dev/null

An http/2 GET is returning an invalid upgrade header (of h2,h2c) which was probably only intended for http/1.1 requests. curl reports it as an error rather more clearly than safari does!

[...]
* Request completely sent off
< HTTP/2 200
< server: nginx/1.24.0 (Ubuntu)
< date: Sat, 22 Aug 2026 11:54:06 GMT
< content-type: image/png
< content-length: 33756
* Invalid HTTP header field was received: frame type: 1, stream: 1, name: [upgrade], value: [h2,h2c]
* [HTTP2] [1] received invalid frame: FRAME[HEADERS, len=273, hend=1, eos=0], error -531: Invalid HTTP header field was received
* HTTP/2 stream 1 reset by curl (error 0xfffffded unknown)

* Connection #0 to host c.tile.opentopomap.org:443 left intact
curl: (92) Invalid HTTP header field was received: frame type: 1, stream: 1, name: [upgrade], value: [h2,h2c]

[hdlnii]

hi @arachsys , what does this mean?

[arachsys]

hi @arachsys , what does this mean?

It just means the [abc].tile.opentopomap.org web server is misconfigured, and the curl output explicitly pinpoints what the problem actually is to make it easy to fix.

[krusty82]

Thank you for figuring out the problem. But why is there nor error for other tiles?
curl --http2 -v https://c.tile.opentopomap.org/8/125/8.png -o /dev/null
For example works.

[arachsys]

Intriguing isn't it? If you make http/1.1 requests to those two URLs, the one that fails is served with Upgrade: h2,h2c (which is fine, but needs to be stripped out in http/2 per rfc7540 8.1.2.2: not doing this is what causes the error above) whereas the one that succeeds is served without the Upgrade: h2,h2c header on both http/1.1 and http/2.

(One mechanism that might cause this is if the front-end nginx reverse-proxies requests to several backend http sources, and different tiles originate from (or are pinned to) different sources. If some (but not all) of those backends have http2 enabled, they might correctly serve the Upgrade: headers, and a naive proxy might forward them intact from an http1.1 backend connection to an http2 client. I've no idea if the tile servers are configured this way, though.)

[arachsys]

I think this problem might now have resolved? I'm no longer able to reproduce it with any of my previously 'bad' examples.

[krusty82]

Cool to hear. I changed some things on the server. But directly after the changes, the error was still there. But maybe the server just needed some time.
Still can't figure out why just some tiles were affected. They all come from the same backend.