inbound fees: collect and display stats #113
Conversation
| Name: inboundFeeBaseName, | ||
| Help: inboundFeeBaseHelp, |
There was a problem hiding this comment.
maybe there's a better way to avoid the constants and extract the name and help text from the GraphCollector instance?
There was a problem hiding this comment.
Ah as attributes? I think the constants fit better since they bind the name, label, etc.
There was a problem hiding this comment.
I was more looking at the pattern that we use for other values
ch <- prometheus.MustNewConstMetric(
g.minTimelockDeltaDesc, prometheus.GaugeValue,
timelockReport.min,
)where we take the description which is defined in a single place in NewGraphCollector, but in this PR the constants are referenced in two places. We could use prometheus.MustNewConstHistogram to follow that pattern, but that seems to be harder to use since we would need to keep track of buckets ourselves.
This lets us collect inbound fee metrics from the graph.
bitromortac
force-pushed
the
2410-inbound-fees
branch
from
127cd89 to
b6c4352
Compare
| Name: inboundFeeBaseName, | ||
| Help: inboundFeeBaseHelp, |
There was a problem hiding this comment.
Ah as attributes? I think the constants fit better since they bind the name, label, etc.
collectors/graph_collector.go
Outdated
| Name: inboundFeeRateName, | ||
| Help: inboundFeeRateHelp, | ||
| Buckets: prometheus.ExponentialBuckets( | ||
| 1, 2, 13, // 1 to 8192 PPM |
There was a problem hiding this comment.
Should we increase this upper range? IIUC this goes up to 0.80% or 80 bps?
There was a problem hiding this comment.
👍 I put some random number here initially, but just checked that the 99-percentile at the moment is 10000 PPM. I bumped it up to an exponent of 15, giving 32768 PPM (~3 %).
We collect non-zero inbound base fees and fee rates in terms of histograms to track the adoption and values set. The histograms are separated into positive and negative portions.
bitromortac
force-pushed
the
2410-inbound-fees
branch
from
b6c4352 to
f7938f7
Compare
|
Thank you for taking a look! Changed that initial range and also cross-checked that the number of negative inbound fee rate policies matches with a manual evaluation. |
|
@Roasbeef: review reminder |
Adds support for tracking the adoption of inbound fees. This is done by collecting non-zero inbound fees based on advertised gossip channel updates using the custom records field in form of histograms for positive and negative values for base fees and fee rates. This lets us analyze negative/positive inbound fee percentiles as well as the overall number of policies that set inbound fees.
Fixes #107 (I have not added total amount of channel updates gauge, since this can be read off from the highest histogram bucket I think, see the dashboard).
Todo: