Timeouts

[angerman]

Hi there! Thanks for building nix-serve-ng. If you permit, let me pick your brains a bit.
With nix-serve-ng, I'm seeing much less of

upstream timed out (110: Connection timed out) while reading response header from upstream

than I saw with nix-serve, where it was almost a constant stream of time outs.

Is there a way nix-serve-ng could report response times per request somehow? I'm not sure I fully trust nginx not to screw something up somewhere in the pipeline.

Another idea I had--though this might be completely misguided--is it possible that the sqlite database has too much contention? I would assume that reads are done concurrently. However if there are writes happening at the same time, and this means the DB blocks a read for a while, this might explain the time outs?

[Gabriella439]

I believe we do support per-request timeouts. See:

nix-serve-ng/src/Options.hs

Lines 132 to 141 in 5e0e4fb

	parseTimeout :: Parser Natural
	parseTimeout =
	Options.option Options.auto
	( Options.long "timeout"
	<> Options.help "Timeout for requests"
	<> Options.metavar "SECONDS"
	-- nix-serve does not timeout requests, but warp insists on a
	-- timeout, so we use the same default timeout as hydra
	<> Options.value (10 * 60)
	)