Open opme opened 1 year ago
Running the full 110K feeds now with this pull request. Database cpu and network is reduced.
Setup is 2 cpu app server with 16 workers, 1 cpu postgres server.
Still want to reduce the spikes.
Feed table has most reads:
current cpu on app server, Want to get it more even without the hills and valleys. It maxes out the 2 cpu for some time then falls to almost zero.
A few thoughts:
Lastly, let me warn you that the latest revision of the code (v0.14) has a new implementation of RSS fetch scheduling, and that it's in its infancy, and not at ALL well tuned, either in terms of database access or in terms of algorithmic complexity (I can point to many places that are likely to have O(n^2) worst case behavior)!!!
improve performance saving stories from feed. In the case where there are 500 stories that need duplicate checking, this would reduce sql to database from 1000 to 2. interested in feedback for this pr. So far tested against the small dokku feedset. I have the full 110k feeds running on a cloud server and will deploy there also.
Todo: add an index to Stories.sources_id batch insert of new stories into stories table?