Closed sharnoff closed 2 months ago
Discussion from the 1-1 meeting: scraping sql_exporter from the autoscaling-agent is the right way to go. I'll add a separate sql_exporter solely for autoscaling metrics. For now, we can just scrape the current one with a reasonable interval.
Discussion with @Omrigan : It's fine to just hard-code the metric names, and that makes it a lot simpler. Some discussion about whether to encapsulate the callbacks for metrics collection into a single struct with func(...)
fields or an interface.
New changes:
4b06d09
..0bb2b59
.metricsMgr[M]
struct to contain them and the kind
string: 5dc98a95675fcb6573278834a688b123eda89149emptyMetrics()
constructor callback to a generics-based approach: 031cc2d13a74e4c232122053404b6119cae04f47
edit: And the first commit has been spun off into a separate PR: #948.
Continuing with the changelog approach, if only because it helps me manage the rebasing:
60f835b
..666e11f
newMetrics
-> updateMetrics
: cb14dafe0dcefdb61bec02e365c9626af81b2cfa+b9e3ebd1ea51a7d3730b965d3ecd77c0318776f4FromPrometheus
: 662613d170363fa9cf240e5a0225e0c2f1c09674c
/l
context/logger variables to ctx2
/logger2
: efdda7c819bbfe1a5178b884d9803438b5cebba1Plus a couple more changes to #948 based on review here.
Final changes:
66b9289
..20bd2ba
20bd2ba
..fb2f910
metricsMgr[M any]
-> metricsMgr[M core.FromPrometheus]
: d0991a9d2447ff1c2a37e0cf2241422df51dea52Waiting on some issues with latest release. Will hold off on merging for now to avoid accidentally adding much larger changes alongside a small fix.
Probably best to review each commit separately.
Most of the commits here are just repositioning to make the final commit, which adds LFC metrics collection, as simple as possible.
Builds on #892, must not be merged before then.