openstreetmap / operations

OSMF Operations Working Group issue tracking
https://operations.osmfoundation.org/
98 stars 13 forks source link

Sporadic issues with S3 minutely replica download fails (pyosmium) #1011

Closed flohoff closed 7 months ago

flohoff commented 7 months ago

Hi, i am running osm pipelines every 30 minutes for different tasks and i update pbfs with /usr/bin/pyosmium-up-to-date from Debian Bookworm.

Since the S3 transition i have failed pipelines about 1-2 times a week.

I fetches replication minutelys - something 20-30 and then suddenly one of them fails. This machine does not have a firewall and is directly connected with a public IP. So i doubt its a client side issue with dropping a tcp/http session. I can also rule out DNS issues as one can see its valid resolvable and previous downloads work.

I am not saying AWS/OSMops is at fault. It might be that sporadic issues on AWS/S3 must be accounted for and pyosmium needs fixing.

[ ... ]
2023-12-06 00:15:26 DEBUG: Downloaded change 5857902. (3266583 kB available in download buffer)
2023-12-06 00:15:26 DEBUG: https://planet.osm.org:443 "GET /replication/minute//005/857/903.osc.gz HTTP/1.1" 302 381
2023-12-06 00:15:26 DEBUG: https://osm-planet-eu-central-1.s3.dualstack.eu-central-1.amazonaws.com:443 "GET /planet/replication/minute/005/857/903.osc.gz HTTP/1.1" 200 49287
2023-12-06 00:15:26 DEBUG: Downloaded change 5857903. (3266370 kB available in download buffer)
2023-12-06 00:15:26 DEBUG: https://planet.osm.org:443 "GET /replication/minute//005/857/904.osc.gz HTTP/1.1" 302 381
2023-12-06 00:15:26 DEBUG: https://osm-planet-eu-central-1.s3.dualstack.eu-central-1.amazonaws.com:443 "GET /planet/replication/minute/005/857/904.osc.gz HTTP/1.1" 200 19641
2023-12-06 00:15:26 DEBUG: Downloaded change 5857904. (3266293 kB available in download buffer)
2023-12-06 00:15:26 DEBUG: Resetting dropped connection: planet.osm.org
2023-12-06 00:15:26 DEBUG: https://planet.osm.org:443 "GET /replication/minute//005/857/905.osc.gz HTTP/1.1" 302 381
2023-12-06 00:15:26 DEBUG: Resetting dropped connection: osm-planet-eu-central-1.s3.dualstack.eu-central-1.amazonaws.com
2023-12-06 00:15:26 DEBUG: Error during diff download. Bailing out.
2023-12-06 00:15:26 DEBUG: https://planet.osm.org:443 "GET /replication/minute//005/857/904.state.txt HTTP/1.1" 302 384
2023-12-06 00:15:26 DEBUG: Starting new HTTPS connection (2): osm-planet-eu-central-1.s3.dualstack.eu-central-1.amazonaws.com:443
2023-12-06 00:15:26 DEBUG: https://osm-planet-eu-central-1.s3.dualstack.eu-central-1.amazonaws.com:443 "GET /planet/replication/minute/005/857/904.state.txt HTTP/1.1" 200 86
2023-12-06 00:15:26 DEBUG: Merging changes into OSM file.
2023-12-06 00:15:47 INFO: Downloaded until 5857904. Server has data available until 5863549.
root@m3 ~ # dpkg -l pyosmium
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name           Version      Architecture Description
+++-==============-============-============-==============================================
ii  pyosmium       3.6.0-1      all          Osmium library bindings for Python - utilities
tomhughes commented 7 months ago

You need to upgrade your pyosmium - https://github.com/osmcode/pyosmium/pull/237

pnorman commented 6 months ago

Duplicate of #992