Hi Steve, some responses from an osmosis point of view below. Steve Hill wrote: > On Tue, 21 Apr 2009, Shaun McDonald wrote: > > >> The renderers are delayed until the minutely diff are working again. >> There are some bugs there that need to be resolved first. Then there >> will be a delay of a few hours/days once the diffs come back up. >> No idea :-) > 2. I noticed the minutely deltas are now delayed by an hour (used to be > about 5 minutes). Is this down to server load? If so, why are the > servers suddenly so much busier than before the API change (there don't > seem to be that many uploads happenning and I wouldn't have thought there > would be a big increase in downloads since the API change). > They shouldn't be delayed. The dev server clock was around 30 minutes slow for a while but this has been fixed. I've now updated osmosis to read the database clock to avoid this happening again. A heavily loaded database server that is still allowing new connections shouldn't slow down minute changesets, it takes just over 2 seconds for a typical changeset to be generated and that includes jvm startup time. If the database server stops accepting new connections (has happened once due to connection limit being exceeded) then changesets will stop completely but will rapidly catch up again when the problem is rectified at the rate of approximately 25 changesets per minute.
I've just checked now and unless I'm confused by timezones it seems to be running 5 minutes behind as per normal. With the current database performing quite well and never causing random query delays with locked tables I may be able to reduce the 5 minute delay down to 1 or 2 but I'm holding off on that one until I'm more sure. > 3. The deltas are much smaller than they used to be and there are long > periods where they are completely empty. A look at the "recent changes" > page seems to show that there really are long periods when no one is > committing any data. Is this down to people actually not trying to upload > anything or is the API undergoing periods of breakage so that people can't > upload anything? > The empty periods correspond to API downtime. I believe there has been issues with API daemons locking up which result in no changes for osmosis to process. As for the size, looking at daily diffs doesn't show any drastic difference in change sizes that can't be explained by the outage periods. Brett _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk