mirrorbits: DB update race

[lazka]

Not that important hopefully, but

• update DB
• user gets redirected to mirror that has update
• user gets redirected to mirror that hasn't and sees pacman downgrade warnings (which aren't fatal)

Workaround:

• force a mirrorbits file rescan after updating the DB (happens all 5 mins anyway)
• this will mark the DBs out of sync with all mirrors and users get redirected to the fallback (repo.msys2.org) until the next mirror update

[lazka]

A docker-compose exec mirrorbits mirrorbits refresh does not seem to help, sometimes

I somehow get the feeling that there is some extra caching on top of redis involved..

[lazka]

A docker-compose exec mirrorbits mirrorbits refresh does not seem to help, sometimes

I somehow get the feeling that there is some extra caching on top of redis involved..

This has been fixed upstream now: etix/mirrorbits#114

So the window is now max 5 minutes, which is really not that much of an issue.

I've filed etix/mirrorbits#170 upstream which would reduce it to 20 seconds.