separates out connection-cache metrics for different protocols #31803
Conversation
behzadnouri
force-pushed
the
connection-cache-metric
branch
from
9663af8
to
3fa54e7
Compare
Codecov Report
@@ Coverage Diff @@
## master #31803 +/- ##
=========================================
- Coverage 81.9% 81.9% -0.1%
=========================================
Files 738 738
Lines 205902 205906 +4
=========================================
- Hits 168705 168662 -43
- Misses 37197 37244 +47 |
There was a problem hiding this comment.
I like that the metrics are grouped now, the one concern beside "bench-tps" naming is that we might end up with too many metrics at the end with some not following naming convention you've introduced implicitly here. One solution might be to use enums instead but this will require bigger boilerplate so don't have a strong opinion
bench-tps/src/main.rs
Outdated
| @@ -83,10 +83,16 @@ fn create_connection_cache( | |||
| client_node_id: Option<&Keypair>, | |||
| ) -> ConnectionCache { | |||
| if !use_quic { | |||
| return ConnectionCache::with_udp(tpu_connection_pool_size); | |||
| return ConnectionCache::with_udp( | |||
| "connection_cache_bench_tps_udp", | |||
There was a problem hiding this comment.
bench-tps metrics are usually named like bench-tps-.*, so it is simpler to search based on the bench client name.
Outside of scope of this PR, would be generally good to agree on some convention about metrics names (snake case vs kebab, start with capital letter or not etc)
There was a problem hiding this comment.
renamed to bench-tps-*
behzadnouri
force-pushed
the
connection-cache-metric
branch
from
3fa54e7
to
6cf31b4
Compare
Problem
Need to distinguish connection-cache metrics for different protocols (e.g. TPU vs turbine).
Summary of Changes
Separated out connection-cache metrics for different protocols