John Spray comments

Results 452 comments of


                                            John Spray

fix(pageserver): limit num of delta layers for l0 compaction

> seems like there are quite a lot test cases asserting on the compaction result layer size/num, probably we should do threshold*checkpoint*2 to avoid hitting those issues I suspect these...

pageserver, safekeeper: make authentication mandatory without dev mode

I cleaned this up manually, Devin was struggling to understand the stack of changes. This ought to start passing tests once today's release's compatibility snapshot is uploaded, so can re-run...

pageserver: exclude un-read layers from short residence statistic

> Generally I'm open to axing a good bunch in the LayerAccessStats Same, I was mulling a really lightweight end state in the future where we might store an `atime`...

safekeeper: block deletion on protocol handler shutdown

Addressed some comments & rebased on main, a few still changes are still oustanding

safekeeper: block deletion on protocol handler shutdown

> One more note: it would be reasonable to add GateGuard to Manager as well, otherwise deletion doesn't wait for manager exit. Added GateGuard in the task where we run...

safekeeper: block deletion on protocol handler shutdown

Sorry for the high-latency turnaround. I've been through all the comments now, although haven't done a full re-self-review to make sure it all still makes sense.

Rename tenant shard request param

Sorry, I missed this one - feel free to bug me if I'm slow on reviews!

control_plane & storage_controller: Request for OpenAPI spec files

It would be fine to open a PR that adds an openAPI description of the storage controller API. Note that the controller effectively has multiple APIs (`/v1/` is a pageserver-compatible...

control_plane & storage_controller: Request for OpenAPI spec files

Looking at the get_pet_by_id example on the utopia homepage, this doesn't look great: it's basically inlining swagger in the Rust code as a macro. which would bloat things quite a...

pageserver: understand resident size spike after deployment + fix it (likely image compaction)

Perhaps something we can observe via existing layer visibility mechanism: we _should_ mark older layers as not-visible when they are covered by image layers, but: - maybe something is broken...