feat!: add summary to coder ping (#14762)

This commit is contained in:
Ethan
2024-09-25 13:24:23 +10:00
committed by GitHub
parent e086d7813b
commit b7c574f679
8 changed files with 258 additions and 93 deletions
+137 -52
View File
@@ -4,26 +4,83 @@ import (
"context"
"errors"
"fmt"
"io"
"net/http"
"net/netip"
"strings"
"time"
"golang.org/x/xerrors"
"tailscale.com/ipn/ipnstate"
"tailscale.com/tailcfg"
"cdr.dev/slog"
"cdr.dev/slog/sloggers/sloghuman"
"github.com/briandowns/spinner"
"github.com/coder/pretty"
"github.com/coder/coder/v2/cli/cliui"
"github.com/coder/coder/v2/cli/cliutil"
"github.com/coder/coder/v2/coderd/util/ptr"
"github.com/coder/coder/v2/codersdk"
"github.com/coder/coder/v2/codersdk/healthsdk"
"github.com/coder/coder/v2/codersdk/workspacesdk"
"github.com/coder/serpent"
)
type pingSummary struct {
Workspace string `table:"workspace,nosort"`
Total int `table:"total"`
Successful int `table:"successful"`
Min *time.Duration `table:"min"`
Avg *time.Duration `table:"avg"`
Max *time.Duration `table:"max"`
Variance *time.Duration `table:"variance"`
latencySum float64
runningAvg float64
m2 float64
}
func (s *pingSummary) addResult(r *ipnstate.PingResult) {
s.Total++
if r == nil || r.Err != "" {
return
}
s.Successful++
if s.Min == nil || r.LatencySeconds < s.Min.Seconds() {
s.Min = ptr.Ref(time.Duration(r.LatencySeconds * float64(time.Second)))
}
if s.Max == nil || r.LatencySeconds > s.Min.Seconds() {
s.Max = ptr.Ref(time.Duration(r.LatencySeconds * float64(time.Second)))
}
s.latencySum += r.LatencySeconds
d := r.LatencySeconds - s.runningAvg
s.runningAvg += d / float64(s.Successful)
d2 := r.LatencySeconds - s.runningAvg
s.m2 += d * d2
}
// Write finalizes the summary and writes it
func (s *pingSummary) Write(w io.Writer) {
if s.Successful > 0 {
s.Avg = ptr.Ref(time.Duration(s.latencySum / float64(s.Successful) * float64(time.Second)))
}
if s.Successful > 1 {
s.Variance = ptr.Ref(time.Duration((s.m2 / float64(s.Successful-1)) * float64(time.Second)))
}
out, err := cliui.DisplayTable([]*pingSummary{s}, "", nil)
if err != nil {
_, _ = fmt.Fprintf(w, "Failed to display ping summary: %v\n", err)
return
}
width := len(strings.Split(out, "\n")[0])
_, _ = fmt.Println(strings.Repeat("-", width))
_, _ = fmt.Fprint(w, out)
}
func (r *RootCmd) ping() *serpent.Command {
var (
pingNum int64
@@ -46,6 +103,14 @@ func (r *RootCmd) ping() *serpent.Command {
ctx, cancel := context.WithCancel(inv.Context())
defer cancel()
spin := spinner.New(spinner.CharSets[5], 100*time.Millisecond)
spin.Writer = inv.Stderr
spin.Suffix = pretty.Sprint(cliui.DefaultStyles.Keyword, " Collecting diagnostics...")
spin.Start()
notifyCtx, notifyCancel := inv.SignalNotifyContext(ctx, StopSignals...)
defer notifyCancel()
workspaceName := inv.Args[0]
_, workspaceAgent, err := getWorkspaceAndAgent(
ctx, inv, client,
@@ -77,11 +142,64 @@ func (r *RootCmd) ping() *serpent.Command {
defer conn.Close()
derpMap := conn.DERPMap()
_ = derpMap
diagCtx, diagCancel := context.WithTimeout(inv.Context(), 30*time.Second)
defer diagCancel()
diags := conn.GetPeerDiagnostics()
// Silent ping to determine whether we should show diags
_, didP2p, _, _ := conn.Ping(ctx)
ni := conn.GetNetInfo()
connDiags := cliui.ConnDiags{
DisableDirect: r.disableDirect,
LocalNetInfo: ni,
Verbose: r.verbose,
PingP2P: didP2p,
TroubleshootingURL: appearanceConfig.DocsURL + "/networking/troubleshooting",
}
awsRanges, err := cliutil.FetchAWSIPRanges(diagCtx, cliutil.AWSIPRangesURL)
if err != nil {
opts.Logger.Debug(inv.Context(), "failed to retrieve AWS IP ranges", slog.Error(err))
}
connDiags.ClientIPIsAWS = isAWSIP(awsRanges, ni)
connInfo, err := wsClient.AgentConnectionInfoGeneric(diagCtx)
if err != nil || connInfo.DERPMap == nil {
return xerrors.Errorf("Failed to retrieve connection info from server: %w\n", err)
}
connDiags.ConnInfo = connInfo
ifReport, err := healthsdk.RunInterfacesReport()
if err == nil {
connDiags.LocalInterfaces = &ifReport
} else {
_, _ = fmt.Fprintf(inv.Stdout, "Failed to retrieve local interfaces report: %v\n", err)
}
agentNetcheck, err := conn.Netcheck(diagCtx)
if err == nil {
connDiags.AgentNetcheck = &agentNetcheck
connDiags.AgentIPIsAWS = isAWSIP(awsRanges, agentNetcheck.NetInfo)
} else {
var sdkErr *codersdk.Error
if errors.As(err, &sdkErr) && sdkErr.StatusCode() == http.StatusNotFound {
_, _ = fmt.Fprint(inv.Stdout, "Could not generate full connection report as the workspace agent is outdated\n")
} else {
_, _ = fmt.Fprintf(inv.Stdout, "Failed to retrieve connection report from agent: %v\n", err)
}
}
spin.Stop()
cliui.PeerDiagnostics(inv.Stderr, diags)
connDiags.Write(inv.Stderr)
results := &pingSummary{
Workspace: workspaceName,
}
n := 0
didP2p := false
start := time.Now()
pingLoop:
for {
if n > 0 {
time.Sleep(pingWait)
@@ -91,6 +209,7 @@ func (r *RootCmd) ping() *serpent.Command {
ctx, cancel := context.WithTimeout(ctx, pingTimeout)
dur, p2p, pong, err := conn.Ping(ctx)
cancel()
results.addResult(pong)
if err != nil {
if xerrors.Is(err, context.DeadlineExceeded) {
_, _ = fmt.Fprintf(inv.Stdout, "ping to %q timed out \n", workspaceName)
@@ -146,57 +265,24 @@ func (r *RootCmd) ping() *serpent.Command {
pretty.Sprint(cliui.DefaultStyles.DateTimeStamp, dur.String()),
)
if n == int(pingNum) {
break
}
}
diagCtx, diagCancel := context.WithTimeout(inv.Context(), 30*time.Second)
defer diagCancel()
diags := conn.GetPeerDiagnostics()
cliui.PeerDiagnostics(inv.Stdout, diags)
ni := conn.GetNetInfo()
connDiags := cliui.ConnDiags{
PingP2P: didP2p,
DisableDirect: r.disableDirect,
LocalNetInfo: ni,
Verbose: r.verbose,
TroubleshootingURL: appearanceConfig.DocsURL + "/networking/troubleshooting",
}
awsRanges, err := cliutil.FetchAWSIPRanges(diagCtx, cliutil.AWSIPRangesURL)
if err != nil {
opts.Logger.Debug(inv.Context(), "failed to retrieve AWS IP ranges", slog.Error(err))
}
connDiags.ClientIPIsAWS = isAWSIP(awsRanges, ni)
connInfo, err := wsClient.AgentConnectionInfoGeneric(diagCtx)
if err != nil || connInfo.DERPMap == nil {
return xerrors.Errorf("Failed to retrieve connection info from server: %w\n", err)
}
connDiags.ConnInfo = connInfo
ifReport, err := healthsdk.RunInterfacesReport()
if err == nil {
connDiags.LocalInterfaces = &ifReport
} else {
_, _ = fmt.Fprintf(inv.Stdout, "Failed to retrieve local interfaces report: %v\n", err)
}
agentNetcheck, err := conn.Netcheck(diagCtx)
if err == nil {
connDiags.AgentNetcheck = &agentNetcheck
connDiags.AgentIPIsAWS = isAWSIP(awsRanges, agentNetcheck.NetInfo)
} else {
var sdkErr *codersdk.Error
if errors.As(err, &sdkErr) && sdkErr.StatusCode() == http.StatusNotFound {
_, _ = fmt.Fprint(inv.Stdout, "Could not generate full connection report as the workspace agent is outdated\n")
} else {
_, _ = fmt.Fprintf(inv.Stdout, "Failed to retrieve connection report from agent: %v\n", err)
select {
case <-notifyCtx.Done():
break pingLoop
default:
if n == int(pingNum) {
break pingLoop
}
}
}
connDiags.Write(inv.Stdout)
if didP2p {
_, _ = fmt.Fprintf(inv.Stderr, "✔ You are connected directly (p2p)\n")
} else {
_, _ = fmt.Fprintf(inv.Stderr, "❗ You are connected via a DERP relay, not directly (p2p)\n%s#common-problems-with-direct-connections\n", connDiags.TroubleshootingURL)
}
results.Write(inv.Stdout)
return nil
},
}
@@ -218,8 +304,7 @@ func (r *RootCmd) ping() *serpent.Command {
{
Flag: "num",
FlagShorthand: "n",
Default: "10",
Description: "Specifies the number of pings to perform.",
Description: "Specifies the number of pings to perform. By default, pings will continue until interrupted.",
Value: serpent.Int64Of(&pingNum),
},
}