Hello,
Are there times of the day when the overpass API https://overpass-api.de/api/interpreter is slower than others ? I get few time outs, but it would be nice if I could avoid the peak usage times - when possible.
Thank you,
Stuart
https://overpass-api.de/munin/localdomain/localhost.localdomain/cpu.html would indicate no real low-load time, but 4am or so is a bit better than the rest.
Simon
Am 09.01.2020 um 20:18 schrieb European Water Project:
Hello,
Are there times of the day when the overpass API https://overpass-api.de/api/interpreter%C2%A0 is slower than others ? I get few time outs, but it would be nice if I could avoid the peak usage times - when possible.
Thank you,
Stuart
talk-ch mailing list talk-ch@openstreetmap.ch http://lists.openstreetmap.ch/mailman/listinfo/talk-ch
Hi Stuart
To add to Simons answer, maybe these graphs are a bit more enlightening: http://overpass-api.de/munin/other-day.html to see the load of the server.
You can also query your own status at https://overpass-api.de/api/status
Generally, you get better success getting a slot on an overcrowded server by making smaller queries. So maybe divide your query into square bboxes and wait some seconds in between until you are granted a new slot.
Hope that helps Michael
Hi Michael and Simon,
Thanks for your replies.
What makes the same overpasa query work at one moment and timeout at another?
Is there a queue system with a max CPU allotted to each query process with a different cpu allocation based on server load?
I essentially do what you are saying by splitting the world into quadrants and looping through.... with a 5 minute delay between queries.
I don't actually make that many queries ~ 200 to get the world fountains...And I have no current plans to run the query more than once a day so for now it's more out of curiosity. ....
Thanks,
Stuart
On Thu, 9 Jan 2020, 20:56 michael spreng, mailinglist@osm.datendelphin.net wrote:
Hi Stuart
To add to Simons answer, maybe these graphs are a bit more enlightening: http://overpass-api.de/munin/other-day.html to see the load of the server.
You can also query your own status at https://overpass-api.de/api/status
Generally, you get better success getting a slot on an overcrowded server by making smaller queries. So maybe divide your query into square bboxes and wait some seconds in between until you are granted a new slot.
Hope that helps Michael
talk-ch mailing list talk-ch@openstreetmap.ch http://lists.openstreetmap.ch/mailman/listinfo/talk-ch
Hi
On 09/01/2020 22:17, European Water Project wrote:
Is there a queue system with a max CPU allotted to each query process with a different cpu allocation based on server load?
Also memory, but yes, that is roughly how it works. See https://overpass-api.de/command_line.html and there section 504 Gateway Timeout. So it depends on you settings timeout and maxsize. If the server is over capacity, it rejects new queries outright if they specify too much time or memory.
Michael
Thank you Michael
How can one have any idea of the maxsize of allowed memory needed for a query on overpass ?
The maximum geojson collection size returned in this project's case is approximately 22MB and this is decompressed. I assume overpass returns a compressed format which should be much smaller.
Best regards,
Stuart
On Fri, 10 Jan 2020 at 06:54, michael spreng < mailinglist@osm.datendelphin.net> wrote:
Hi
On 09/01/2020 22:17, European Water Project wrote:
Is there a queue system with a max CPU allotted to each query process with a different cpu allocation based on server load?
Also memory, but yes, that is roughly how it works. See https://overpass-api.de/command_line.html and there section 504 Gateway Timeout. So it depends on you settings timeout and maxsize. If the server is over capacity, it rejects new queries outright if they specify too much time or memory.
Michael _______________________________________________ talk-ch mailing list talk-ch@openstreetmap.ch http://lists.openstreetmap.ch/mailman/listinfo/talk-ch
Hi
On 10/01/2020 07:20, European Water Project wrote:
How can one have any idea of the maxsize of allowed memory needed for a query on overpass ?
I think the only way is to reduce time or memory until you get an error. Then you know roughly how much you need. Memory means ram used for the query, not how large the response is. You don't have to get those two numbers exactly right, it just helps if they are not orders of magnitude off for your chances to get a slot on overpass. And to catch errors and increase those numbers when needed because otherwise you won't get a response :)
Michael
-
European Water Project -
michael spreng -
Simon Poole