Pending AIs
[bmbouter] open proposal issue to remove queues
[fabricio/dennis] ask katello if they want ruby bindings docs
[brian/matthias] open django prefetch bug
March 22
- memory PoC continued!
- 27% reduction in memory by reducing queue count from 1000 to 1
- should we remove the queues entirely?
- concern: deadlocking due first-stage waiting on the resolution of a future, or some even the pipeline is supposed to produce
- more of a concern now that the pipeline is even less filled
- 27% reduction in memory by reducing queue count from 1000 to 1
- proposal: removing django admin UI views
- bindings docs
- should we publish for both: ruby and python?
- https://github.com/pulp/plugin_template/pull/599
- let’s ask katello
- mkdocs switch?
- should we?
- fabricio will show his PoC
- weird “db connection already closed” CI failure https://github.com/pulp/pulpcore/pull/2357#issuecomment-1072333934
- maybe issue with a test, dkliban will take a look
- absolutely weird RemoteArtifactSaver failure in pulp_container
- Goal: Deduplicating tests between pulpcore and pulp_file
March 15th
- memory PoC concluded
- adjusting batch size is effective at reducing memory
- see the results
- dynamically adjusting it doesn’t keep it within a “target amount”
- it runs extremely slow 10x plus
- memory is never returned to the OS, the pages mostly just stay allocated and later
- new memory charting debugging feature added:
- proposal: to expose the batch size to users, the batch size is likely the single best adjustment to reduce memory
- we need each plugin to also look at their own first-stage code to see if it can be memory optimized
- proposal: https://github.com/pulp/pulpcore/issues/2069#issuecomment-1066853598
- adjusting batch size is effective at reducing memory
- maintainability policy
- how far should we keep ensuring old branches are working?
- single container currently does latest 3 y-stream
- pulp-cli tests with 5 y-streams