[*] optimize flush() in pg sink

[0xgouda]

We don't need to loop over all the rows in Measurements and read epoch_ns from each, because they typically all contain the same value extracted from now() in the same txn.

[coveralls]

Coverage Report for CI Build 33176349266

Coverage decreased (-0.1%) to 87.878%

Details

• Coverage decreased (-0.1%) from the base build.
• Patch coverage: 5 of 5 lines across 2 files are fully covered (100%).
• 8 coverage regressions across 1 file.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

8 previously-covered lines in 1 file lost coverage.

File	Lines Losing Coverage	Coverage
internal/sinks/prometheus.go	8	93.56%

---

Coverage Stats

Relevant Lines:	6154
Covered Lines:	5408
Line Coverage:	87.88%
Coverage Strength:	1.0 hits per line

---

💛 - Coveralls

[pashagolub]

cool! do you have numbers on speed imrovements?

[0xgouda]

cool! do you have numbers on speed imrovements?

Nope, I didn't benchmark it, but I would expect an improvement only for large setups with 100s of monitored sources, with each of them returning a lot of rows.

IMO, smaller setups won't be affected that much.

[pashagolub]

LGTM! Good job!

[0xgouda]

cool! do you have numbers on speed imrovements?

I did a small benchmark on my local machine mimicking 1000 sources, each with 80 active metrics with the same interval, and each metric returns 200 rows, and it was 7% faster:

goos: linux
goarch: amd64
pkg: github.com/cybertec-postgresql/pgwatch/v6/internal/sinks
cpu: Intel(R) Core(TM) i5-10500H CPU @ 2.50GHz
               │  old.txt   │              new.txt              │
               │   sec/op   │   sec/op    vs base               │
PgSinkflush-12   131.0 ± 7%   121.9 ± 4%  -6.92% (p=0.000 n=10)