Think about long running performance and stability tests
To improve stability we should get more insight of the long-term behavior of Akka.
- How long can we afford to run tests?
- Do we want to run these tests for different versions, or the latest, or latest released? Periodic snapshots of master? (two-week, or such)
- Ideal would be tests that run "infinitely long", and have weekly automatic reports (some leaks might materialize as a very slow drift of memory usage so they don't result in crashes).
- Also, a realistic and sufficiently complex system is needed -- a simple ping-pong might not do the trick.
- A good opportunity to dogfood Atmos
Let's Discuss!
- How long can we afford to run tests?
- Do we want to run these tests for different versions, or the latest, or latest released? Periodic snapshots of master? (two-week, or such)
- Ideal would be tests that run "infinitely long", and have weekly automatic reports (some leaks might materialize as a very slow drift of memory usage so they don't result in crashes).
- Also, a realistic and sufficiently complex system is needed -- a simple ping-pong might not do the trick.
- A good opportunity to dogfood Atmos
Let's Discuss!
Leave a comment
What more thinking is needed here? Can we create more concrete tickets if something is not covered by existing tickets?