On 12/01/2009 12:40, Frederik Ramm wrote: > Hi, > > David Earl wrote: >> suspect his vicious circle started because there was one day when >> there was an exceptionally large update and this ran fro > 24 hours, >> running into the next. > > It could also be the fact that there was no daily diff in the early > hours of 1st Jan because the server had frozen; the > 2008-12-13-to-2009-01-01 diff was then generated in the next daily diff > run, 24 hours late, and just an hour later the 01-01-to-01-02 diff came in. > > Maybe in situations like these you can just stop the update and when the > next weekly update comes, use that to re-sync?
I think it was having problems before that, but it may well have made it worse. A missing update shouldn't break it - it just does nothing if nothing is available, and the next night it picks more than one up. It runs in two phases - read all unprocessed updates and then update the combined result. It is a little cheaper to run two updates together than each in succession, though it can eat into the European daytime when that happens. In the short term what I need here is a queue rather than a series of blind daily cron job. Not hard to do. David _______________________________________________ dev mailing list [email protected] http://lists.openstreetmap.org/listinfo/dev
