Overpass API developpement

[Igor Brejc <>]

Hi, 

There has been a "remote debugging session" back-and-forth correspondence between Roland and me. I'm not sure what the final outcome of it was, but from what I can gather I don't think it's an issue with a corrupted clone, more like some file permissions issue - but of course Roland should be the judge here.

It could be the problem is related to my (special) workflow - no attic data and no metadata. Given that I managed to reproduce it twice on different distributions/flavors (first Ubuntu 32bit, then XUbuntu 64bit) with a clean machine (and not running inside Docker, just to make this clear) and with Overpass DB clones of different timestamps, I don't think it should be difficult to reproduce it by others (I've documented my steps on https://wiki.openstreetmap.org/wiki/User:Breki/Overpass_API_Installation).

In the meantime I abandoned the idea of having live DB updates of my Overpass instance - I need a functioning service with at least some data, even if it is not the latest one.

Cheers,

Igor

On Mon, Aug 28, 2017 at 7:04 PM, mmd <> wrote:

Am 31.07.2017 um 16:38 schrieb Roland Olbricht:

>
> Please send again the full debugging session including the output from
> the program - it is by default interspersed with the gdb output.
>

It's been quite a while since the last update. Has there been some
progress on this topic? Did the issue reoccur?

Can we make sure that the download itself didn't introduce any file
corruption, which we were chasing all the time? For this purpose there's
a new github issue to have an external checksum for every single clone
file in place, see https://github.com/drolbr/Overpass-API/issues/420.

btw: Based on a May 31st clone over at dev.overpass-api.de, I've created
a new attic database including meta data, converted it to lz4 based
compression and started applying minutely diffs. None of the issues
mentioned in this thread occurred here even after applying almost 3
months worth of minutely diffs.

--