DM-45908: Fix client-side HTTP timeouts when communicating with Butler server #1065
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
dhirving
force-pushed
the
tickets/DM-45908
branch
from
6b79966
to
b0fab7c
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #1065 +/- ##
=======================================
Coverage 89.54% 89.54%
=======================================
Files 359 359
Lines 46473 46509 +36
Branches 9558 9566 +8
=======================================
+ Hits 41612 41648 +36
Misses 3511 3511
Partials 1350 1350 ☔ View full report in Codecov by Sentry. |
The default HTTPX timeout of 5 seconds is much too low when communicating with Butler server -- many requests can exceed that while waiting for the database to respond.
Fetching database results for Butler queries can take an unpredictable amount of time, in some cases several minutes. To ensure that clients do not time out while waiting, send a keep-alive message to the stream every fifteen seconds.
Remove the GzipMiddleware for two reasons: 1. It was preventing "keep-alive" messages from working in streamed responses, because it was batching them until it had a full gzip chunk. 2. It was slowing down queries by about 10% at the RSP -- it turns out that compressing/decompressing is expensive, and there is lots of bandwidth available within Google.
dhirving
force-pushed
the
tickets/DM-45908
branch
2 times, most recently
from
0c1baf5
to
3d29584
Compare
dhirving
commented
Aug 30, 2024
| @@ -54,7 +53,6 @@ def create_app() -> FastAPI: | |||
| config = load_config() | |||
|
|
|||
| app = FastAPI() | |||
| app.add_middleware(GZipMiddleware, minimum_size=1000) | |||
There was a problem hiding this comment.
This was preventing the keep-alives from working because it would buffer/compress them and never send them.
I tested it and it turns out that gzipping was slowing down queries by 10% on the RSP -- there's oodles of bandwidth there and apparently the compression overhead is non-trivial. So I just removed this, since there are no configuration options to make it only apply to certain content-types or routes.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixed an issue where slow queries would result in
httpxtimeouts on the client side. The server now sends keep-alives from slow-running queries to prevent a timeout. The timeout duration has been increased to 2 minutes on the client side -- previously it was 5 seconds.Checklist
doc/changesconfigs/old_dimensions