Closed FANNG1 closed 5 months ago
What's the purpose of writing metrics to FS, not db?
What's the purpose of writing metrics to FS, not db?
Analyzing the metrics with Spark or Python is easy if writing to FS.
remove the specified storage implement, leaving the basic framework.
@Clearvive @jerryshao , could you help to review this pr?
Do we need a metrics storage for the Gravitino server?
Do we need a metrics storage for the Gravitino server?
no, it's used to store Iceberg query or commit metrics not server metrics.
@Clearvive @yuqi1129 , are comments addressed, please review again. thx.
LGTM except the few comments
@jerryshao , all comments are addressed, please help to review again.
planing to do:
planing to do:
- remove async mode configuration
- add queue capacity configutation
- use service loader to load specified metrics storage.
supported in the lastest pr , cc @jerryshao
I think there's no need to use service loader to make it pluggable, using reflection should be enough.
I think there's no need to use service loader to make it pluggable, using reflection should be enough.
replace service loader with reflection.
What changes were proposed in this pull request?
MetricsManager
to manage storagesIcebergMetricsFormatter
to format metrics, support JSON for now.Why are the changes needed?
Fix: #1187
Does this PR introduce any user-facing change?
no
How was this patch tested?