Duplicate nodes in diff

[joto]

See report here https://community.openstreetmap.org/t/duplicate-nodes-in-osc-downloaded-from-osm/117107

See v21 for node 10767916505 and v13 for node 10824026405 are twice in https://planet.openstreetmap.org/replication/day/000/004/346.osc.gz .

The nodes are in two minutely diffs: https://planet.osm.org/replication/minute/006/205/197.osc.gz and https://planet.osm.org/replication/minute/006/205/198.osc.gz . This should not happen. What is conspicuous is that around that time there were no minutely diffs for a while.

198.state.txt            2024-08-05 07:49   86   
198.osc.gz               2024-08-05 07:49  2.5M  
197.state.txt            2024-08-05 06:36   86   
197.osc.gz               2024-08-05 06:36  805   
196.state.txt            2024-08-05 06:35   86   
196.osc.gz               2024-08-05 06:35  851   
195.state.txt            2024-08-05 02:24   86   
195.osc.gz               2024-08-05 02:24  8.7K 

[tomhughes]

Well yes that's when the site was down due to a database issue.

[Slpy-API]

I'm assuming users should fix the file manually, but would appreciate an update beforehand on whether the issue will be fixed in the stream or if any other checks need to be done first. Thanks!

[tomhughes]

Well if somebody can confirm what actually needs to be done then maybe we can do something...

Is the whole of 197 duplicated in 198 or just some of it?

[tomhughes]

Oh in fact those are the only two changes in 197 so it looks like osm-repl-2024-08-05T06:35:46Z-lsn-124BF-C70BD160.log got included in both.

I know I had to kill the minutely replication because it go wedged when the database machine had to be restarted but that kill happened around the time of 198 and 197 was uploaded an hour before that which means that segment was finished and should have been renamed so I don't really understand how it got repeated. There is an odd delay of several minutes between the location version of 197 and the S3 version though, suggesting that the osmdbt-create-diff was quite slow to exit after generating the file.

[tomhughes]

I've removed the duplicates from the minutely, hourly and daily diffs now.