Closed rwst closed 3 years ago
Setting a delay would still allow some requests to fail if the delay becomes too short. My preferred approach would rather be to retry after a delay after having encountered a failed-save, just like when we get a maxlag error: that would allow to adapt to the server capacity without having to guess the appropriate delay.
I just published v15.15.1
, which should do that retry on failed-save
; could you test to see if that solves the problem?
This works fine, many thanks! The fail phases usually are shorter than 30s.
The behaviour of the Wikidata API changed, as seen at https://phabricator.wikimedia.org/T256386 I think adding a delay option (multiples of 0.1s) would allow me to avoid failed-saves.
But I could be wrong about it all. What do you think?