Files
tendermint/proxy/app_conn.go
William Banfield 935fbd8234 internal/proxy: add initial set of abci metrics (#7115)
This PR adds an initial set of metrics for use ABCI. The initial metrics enable the calculation of timing histograms and call counts for each of the ABCI methods. The metrics are also labeled as either 'sync' or 'async' to determine if the method call was performed using ABCI's `*Async` methods.

An example of these metrics is included here for reference:
```
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="0.0001"} 0
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="0.0004"} 5
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="0.002"} 12
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="0.009"} 13
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="0.02"} 13
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="0.1"} 13
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="0.65"} 13
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="2"} 13
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="6"} 13
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="25"} 13
tendermint_abci_connection_method_timing_bucket{chain_id="ci",method="commit",type="sync",le="+Inf"} 13
tendermint_abci_connection_method_timing_sum{chain_id="ci",method="commit",type="sync"} 0.007802058000000001
tendermint_abci_connection_method_timing_count{chain_id="ci",method="commit",type="sync"} 13
```

These metrics can easily be graphed using prometheus's `histogram_quantile(...)` method to pick out a particular quantile to graph or examine. I chose buckets that were somewhat of an estimate of expected range of times for ABCI operations. They start at .0001 seconds and range to 25 seconds. The hope is that this range captures enough possible times to be useful for us and operators.
2022-08-04 15:48:28 -04:00

232 lines
7.7 KiB
Go

package proxy
import (
"time"
"github.com/go-kit/kit/metrics"
abcicli "github.com/tendermint/tendermint/abci/client"
"github.com/tendermint/tendermint/abci/types"
)
//go:generate ../scripts/mockery_generate.sh AppConnConsensus|AppConnMempool|AppConnQuery|AppConnSnapshot
//----------------------------------------------------------------------------------------
// Enforce which abci msgs can be sent on a connection at the type level
type AppConnConsensus interface {
SetResponseCallback(abcicli.Callback)
Error() error
InitChainSync(types.RequestInitChain) (*types.ResponseInitChain, error)
BeginBlockSync(types.RequestBeginBlock) (*types.ResponseBeginBlock, error)
DeliverTxAsync(types.RequestDeliverTx) *abcicli.ReqRes
EndBlockSync(types.RequestEndBlock) (*types.ResponseEndBlock, error)
CommitSync() (*types.ResponseCommit, error)
}
type AppConnMempool interface {
SetResponseCallback(abcicli.Callback)
Error() error
CheckTxAsync(types.RequestCheckTx) *abcicli.ReqRes
CheckTxSync(types.RequestCheckTx) (*types.ResponseCheckTx, error)
FlushAsync() *abcicli.ReqRes
FlushSync() error
}
type AppConnQuery interface {
Error() error
EchoSync(string) (*types.ResponseEcho, error)
InfoSync(types.RequestInfo) (*types.ResponseInfo, error)
QuerySync(types.RequestQuery) (*types.ResponseQuery, error)
// SetOptionSync(key string, value string) (res types.Result)
}
type AppConnSnapshot interface {
Error() error
ListSnapshotsSync(types.RequestListSnapshots) (*types.ResponseListSnapshots, error)
OfferSnapshotSync(types.RequestOfferSnapshot) (*types.ResponseOfferSnapshot, error)
LoadSnapshotChunkSync(types.RequestLoadSnapshotChunk) (*types.ResponseLoadSnapshotChunk, error)
ApplySnapshotChunkSync(types.RequestApplySnapshotChunk) (*types.ResponseApplySnapshotChunk, error)
}
//-----------------------------------------------------------------------------------------
// Implements AppConnConsensus (subset of abcicli.Client)
type appConnConsensus struct {
metrics *Metrics
appConn abcicli.Client
}
func NewAppConnConsensus(appConn abcicli.Client, metrics *Metrics) AppConnConsensus {
return &appConnConsensus{
metrics: metrics,
appConn: appConn,
}
}
func (app *appConnConsensus) SetResponseCallback(cb abcicli.Callback) {
app.appConn.SetResponseCallback(cb)
}
func (app *appConnConsensus) Error() error {
return app.appConn.Error()
}
func (app *appConnConsensus) InitChainSync(req types.RequestInitChain) (*types.ResponseInitChain, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "init_chain", "type", "sync"))()
return app.appConn.InitChainSync(req)
}
func (app *appConnConsensus) BeginBlockSync(req types.RequestBeginBlock) (*types.ResponseBeginBlock, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "begin_block", "type", "sync"))()
return app.appConn.BeginBlockSync(req)
}
func (app *appConnConsensus) DeliverTxAsync(req types.RequestDeliverTx) *abcicli.ReqRes {
defer addTimeSample(app.metrics.MethodTiming.With("method", "deliver_tx", "type", "async"))()
return app.appConn.DeliverTxAsync(req)
}
func (app *appConnConsensus) EndBlockSync(req types.RequestEndBlock) (*types.ResponseEndBlock, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "end_block", "type", "sync"))()
return app.appConn.EndBlockSync(req)
}
func (app *appConnConsensus) CommitSync() (*types.ResponseCommit, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "commit", "type", "sync"))()
return app.appConn.CommitSync()
}
//------------------------------------------------
// Implements AppConnMempool (subset of abcicli.Client)
type appConnMempool struct {
metrics *Metrics
appConn abcicli.Client
}
func NewAppConnMempool(appConn abcicli.Client, metrics *Metrics) AppConnMempool {
return &appConnMempool{
metrics: metrics,
appConn: appConn,
}
}
func (app *appConnMempool) SetResponseCallback(cb abcicli.Callback) {
app.appConn.SetResponseCallback(cb)
}
func (app *appConnMempool) Error() error {
return app.appConn.Error()
}
func (app *appConnMempool) FlushAsync() *abcicli.ReqRes {
defer addTimeSample(app.metrics.MethodTiming.With("method", "flush", "type", "async"))()
return app.appConn.FlushAsync()
}
func (app *appConnMempool) FlushSync() error {
defer addTimeSample(app.metrics.MethodTiming.With("method", "flush", "type", "sync"))()
return app.appConn.FlushSync()
}
func (app *appConnMempool) CheckTxAsync(req types.RequestCheckTx) *abcicli.ReqRes {
defer addTimeSample(app.metrics.MethodTiming.With("method", "check_tx", "type", "async"))()
return app.appConn.CheckTxAsync(req)
}
func (app *appConnMempool) CheckTxSync(req types.RequestCheckTx) (*types.ResponseCheckTx, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "check_tx", "type", "sync"))()
return app.appConn.CheckTxSync(req)
}
//------------------------------------------------
// Implements AppConnQuery (subset of abcicli.Client)
type appConnQuery struct {
metrics *Metrics
appConn abcicli.Client
}
func NewAppConnQuery(appConn abcicli.Client, metrics *Metrics) AppConnQuery {
return &appConnQuery{
metrics: metrics,
appConn: appConn,
}
}
func (app *appConnQuery) Error() error {
return app.appConn.Error()
}
func (app *appConnQuery) EchoSync(msg string) (*types.ResponseEcho, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "echo", "type", "sync"))()
return app.appConn.EchoSync(msg)
}
func (app *appConnQuery) InfoSync(req types.RequestInfo) (*types.ResponseInfo, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "info", "type", "sync"))()
return app.appConn.InfoSync(req)
}
func (app *appConnQuery) QuerySync(reqQuery types.RequestQuery) (*types.ResponseQuery, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "query", "type", "sync"))()
return app.appConn.QuerySync(reqQuery)
}
//------------------------------------------------
// Implements AppConnSnapshot (subset of abcicli.Client)
type appConnSnapshot struct {
metrics *Metrics
appConn abcicli.Client
}
func NewAppConnSnapshot(appConn abcicli.Client, metrics *Metrics) AppConnSnapshot {
return &appConnSnapshot{
metrics: metrics,
appConn: appConn,
}
}
func (app *appConnSnapshot) Error() error {
return app.appConn.Error()
}
func (app *appConnSnapshot) ListSnapshotsSync(req types.RequestListSnapshots) (*types.ResponseListSnapshots, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "list_snapshots", "type", "sync"))()
return app.appConn.ListSnapshotsSync(req)
}
func (app *appConnSnapshot) OfferSnapshotSync(req types.RequestOfferSnapshot) (*types.ResponseOfferSnapshot, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "offer_snapshot", "type", "sync"))()
return app.appConn.OfferSnapshotSync(req)
}
func (app *appConnSnapshot) LoadSnapshotChunkSync(
req types.RequestLoadSnapshotChunk) (*types.ResponseLoadSnapshotChunk, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "load_snapshot_chunk", "type", "sync"))()
return app.appConn.LoadSnapshotChunkSync(req)
}
func (app *appConnSnapshot) ApplySnapshotChunkSync(
req types.RequestApplySnapshotChunk) (*types.ResponseApplySnapshotChunk, error) {
defer addTimeSample(app.metrics.MethodTiming.With("method", "apply_snapshot_chunk", "type", "sync"))()
return app.appConn.ApplySnapshotChunkSync(req)
}
// addTimeSample returns a function that, when called, adds an observation to m.
// The observation added to m is the number of seconds ellapsed since addTimeSample
// was initially called. addTimeSample is meant to be called in a defer to calculate
// the amount of time a function takes to complete.
func addTimeSample(m metrics.Histogram) func() {
start := time.Now()
return func() { m.Observe(time.Since(start).Seconds()) }
}