[blob-service] Implement download logic with new db
ClosedPublic
Actions

Authored by bartek on Jul 9 2023, 2:05 PM.

Details

Reviewers

michal
varun
• patryk
• jon

Commits

rCOMM0c2ea1abaaa5: [blob-service] Implement download logic with new db

Summary

Part of ENG-4269.

Most of this code is copy-pasted from http::handlers::blob::get_blob_handler

Decided to create a separate download object, similar to "upload sessions" we already have. The reasoning for this is that we typically have the following workflow:

Get the download object metadata (size etc.)
Set download params (range, format, streaming options)
Stream the data to the client

Depends on D8453.

Test Plan

Tested later in the stack

Diff Detail

Repository

rCOMM Comm

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

bartek created this revision.Jul 9 2023, 2:05 PM

bartek held this revision as a draft.

Herald added subscribers: tomek, ashoat. · View Herald TranscriptJul 9 2023, 2:05 PM

bartek added a child revision: D8455: [blob-service] Implement upload logic with new db.Jul 9 2023, 2:05 PM

Harbormaster completed remote builds in B20823: Diff 28510.Jul 9 2023, 2:36 PM

bartek edited the summary of this revision. (Show Details)Jul 9 2023, 11:38 PM

bartek edited the test plan for this revision. (Show Details)

bartek edited the summary of this revision. (Show Details)

bartek published this revision for review.Jul 10 2023, 12:43 AM

rust LGTM, would like for someone else to take a look

michal added inline comments.Jul 17 2023, 1:31 AM

services/blob/src/service.rs
78 ↗	(On Diff #28510)	nit: metadata
121 ↗	(On Diff #28510)	Any reason for not using non-inclusive range? Seems like it would simplify the code a bit
154 ↗	(On Diff #28510)	I think this might skip the last byte in some cases (e.g. when `chunk_size == 1` or when `offset` is equal to the byte one before last), but I could be mistaken

• patryk requested changes to this revision.Jul 17 2023, 11:48 PM

• patryk added inline comments.

services/blob/src/service.rs
121 ↗	(On Diff #28510)	Maybe it's because HTTP Range header is inclusive too, but might be another reason.
154 ↗	(On Diff #28510)	Good catch! It skips last byte when `(last_byte - first_byte) % chunk_size == 0`. Changing condition to `offset <= last_byte` should solve the problem.

This revision now requires changes to proceed.Jul 17 2023, 11:48 PM

bartek added inline comments.Jul 19 2023, 3:53 AM

services/blob/src/service.rs
121 ↗	(On Diff #28510)	No opinion on this. Inclusive ranges seemed just more natural and re-used existing code. I can change to non-inclusive
154 ↗	(On Diff #28510)	Yeah, good catch. But it means that previous code was also wrong (this part was copy-pasted) - see D8458