[Israel Fruchter, Test Podcast] (or at least didn't shared they tests, or test plans :))
[Dima Spivak, Test Podcast] you'd obviously want to do things with parallelism to make sure stuff doesn't fall over with many users or whatnot
[Dima Spivak, Test Podcast] but maybe give us a sense of what you're thinking and we can critique it?
[Israel Fruchter, Test Podcast] basically it a implementation of this exact API ontop of out dynamodb drop in replacement.
[Dima Spivak, Test Podcast] yep, got that part. I mean what sort of cases or scenarios do you have in mind to cover?
[Israel Fruchter, Test Podcast] and out plan is to write a tool using https://github.com/awslabs/dynamodb-streams-kinesis-adapter, to duplicate the a table in realtime
[Israel Fruchter, Test Podcast] while using YCSB to populated that table.
[Dima Spivak, Test Podcast] gotcha. so that's more of a load test?
[Israel Fruchter, Test Podcast] and then we could compare both, source and target versions
[Israel Fruchter, Test Podcast] while we are stopping/killing nodes of our DB
[Israel Fruchter, Test Podcast] we also have more specific functional tests
[Dima Spivak, Test Podcast] yep, all seems reasonable. so this is less about testing the API than testing the behavior of your database?
[Israel Fruchter, Test Podcast] that would cover the APIs
[Dima Spivak, Test Podcast] (when I see "API test" I imagine more things around correctness and error handling than load in the face of fault injection)
[Israel Fruchter, Test Podcast] I'm planning both testing the APIs in a more controlled and specific environment
[Israel Fruchter, Test Podcast] but also in a more close to reality environment, with bigger amount of traffic going into the system
[Dima Spivak, Test Podcast] gotcha
[Dima Spivak, Test Podcast] so yeah, for APIs, the things most people forget are around malformed data
[Dima Spivak, Test Podcast] payloads that are too big
[Dima Spivak, Test Podcast] wrong datatypes being passed
[Dima Spivak, Test Podcast] that sorta thing
[Israel Fruchter, Test Podcast] FYI, those are the "unit-tests" the development guys were doing so far
[Israel Fruchter, Test Podcast] https://github.com/scylladb/scylla/blob/master/test/alternator/test_streams.py
[Dima Spivak, Test Podcast] oh man, so nice to see pretty pytest work =D
[Israel Fruchter, Test Podcast] yeah, it is
[Israel Fruchter, Test Podcast] it's not mine, but I do approve most of it :)
[Israel Fruchter, Test Podcast] whom wrote it, it's the first time he's touching pytest (or python for that matter)...
[Israel Fruchter, Test Podcast] but he's really experienced developer, like a 20x one :)
[Dima Spivak, Test Podcast] there's that muscle memory of having read pep-8 90 times that shows in code looking "pythonic"
[Israel Fruchter, Test Podcast] his day job, he's doing C++... so yeah :)
[Israel Fruchter, Test Podcast] So one take I have from you @dima is approaching from the angle of a "naive" developer angle. (and make it "idiot" proof) :)
[Dima Spivak, Test Podcast] absolutely
[Dima Spivak, Test Podcast] the most common issues that surface are from people who don't rtfm
[Dima Spivak, Test Podcast] so having your application respond to that gracefully is important
[Israel Fruchter, Test Podcast] yeah, seen that sentence today on a slide, "we should invest more in helping our users, to no shoot themselves in the leg"
[Dima Spivak, Test Podcast] e.g. people who pass payloads when the endpoint doesn't support it, query parameters when path parameters are expected, etc.
[Israel Fruchter, Test Podcast] we mainly going to use boto3 in the functional tests, kind of assuming most folks won't use HTTP directly with an AWS like API
[Dima Spivak, Test Podcast] a lot of folks still try to use curl for all their automation needs
[Israel Fruchter, Test Podcast] with AWS ?
[Dima Spivak, Test Podcast] ooh yeah.
[Dima Spivak, Test Podcast] lol
[Israel Fruchter, Test Podcast] that the "crazy" developer angle :)
[Dima Spivak, Test Podcast] the things I've seen would give you nightmares
[Dima Spivak, Test Podcast] I mean think about it, the users of your products aren't technical enough to build CassandraDB in C++ themselves
[Dima Spivak, Test Podcast] so they find a drop in replacement
[Dima Spivak, Test Podcast] and then if they know about bash scripts and using curl, they'll keep using that