Experimental: when a scheduled update runs, skip looking for and removing...
Experimental: when a scheduled update runs, skip looking for and removing other scheduled updates for the same feed, before scheduling the next update. My profiling seems to indicate that searching...
View ArticleAdded ruby-prof gem to the bundle. It is used for profiling.
More details: https://github.com/ruby-prof/ruby-prof
View ArticleMaximum update interval changed from 6 to 12 hours. Minimum changed from 10 t...
Less frequent updates, but still should be good enough. This will lighten the load on the background server a bit.
View ArticleResolved relative asset paths for bower components.
Should solve issue with hopscotch images being requested from the wrong path.
View ArticleFixed tests. Tests now use the configured max and min update intervals,...
Fixed tests. Tests now use the configured max and min update intervals, instead of hardcoded values. This means that if the max/min values in the config are changed, the tests will still pass.
View ArticlePerformance improvement when fixing update schedules.
The sidekiq objects for the different named queues are retrieved just once, instead of once for every feed in the db. This should reduce the time obtaining a connection, hopefully reduce...
View ArticlePerformance improvement when fixing schedules.
Querying Redis for each feed is very slow. What we do instead is to get all IDs passed to scheduled updates, all at once for each queue. We have then an array of IDs in a ruby variable for each named...
View ArticleRewrite recurrence of UpdateRedisCacheStatsWorker.
This change is in line with recommendations from the Sidetiq guys: https://github.com/tobiassvn/sidetiq/wiki/Known-Issues#ice_cube Instead of minutely(10) to run the job every 10 minutes, they suggest...
View ArticleUpdate interfal limits are again 6 hours - 15 minutes.
I think I've solved the most important performance problems in the background server. These lower interval limits should be no problem.
View ArticleMark feeds as permanently unavailable after 1 month, not 1 week.
I'm seeing too many feeds become unavailable for more than a week, only to be available again after some time. 1 month is maybe more reasonable.
View ArticleFixed tests.
Tests now use the configured maximum time that a feed can be failing before being marked as permanently unavailable, instead of a hardcoded value of 1 week. This way tests will pass even if this value...
View ArticleWhen fixing schedules, make sure that there is exactly one scheduled update...
When fixing schedules, make sure that there is exactly one scheduled update for each available feed. FixSchedulesWorker now not only adds missing scheduled updates. It also counts how many times are...
View Article