synced/coder
1
0
Fork 0
mirror of https://github.com/coder/coder.git synced 2025-07-12 00:14:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

chore: refactor sending telemetry (#15345)

Browse Source 
Implements a tailnet API Telemetry controller by refactoring from `workspacesdk`.

chore re: #14729
This commit is contained in:
Spike Curtis
2024-11-06 20:23:23 +04:00
committed by GitHub
parent 9126cd78a6
commit 335e4ab6bf
4 changed files with 254 additions and 186 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
codersdk/workspacesdk/connector.go
View File

@ -8,16 +8,12 @@ import (
"net/http"
"net/url"
"slices"
"strings"
"sync"
"sync/atomic"
"time"
"github.com/google/uuid"
"golang.org/x/xerrors"
"nhooyr.io/websocket"
"storj.io/drpc"
"storj.io/drpc/drpcerr"
"cdr.dev/slog"
"github.com/coder/coder/v2/buildinfo"
@ -66,19 +62,12 @@ type tailnetAPIConnector struct {
dialOptions *websocket.DialOptions
derpCtrl tailnet.DERPController
coordCtrl tailnet.CoordinationController
customDialFn func() (proto.DRPCTailnetClient, error)
clientMu sync.RWMutex
client proto.DRPCTailnetClient
telCtrl *tailnet.BasicTelemetryController
connected chan error
resumeToken *proto.RefreshResumeTokenResponse
isFirst bool
closed chan struct{}
// Only set to true if we get a response from the server that it doesn't support
// network telemetry.
telemetryUnavailable atomic.Bool
}
// Create a new tailnetAPIConnector without running it
@ -92,6 +81,7 @@ func newTailnetAPIConnector(ctx context.Context, logger slog.Logger, agentID uui
dialOptions: dialOptions,
connected: make(chan error, 1),
closed: make(chan struct{}),
telCtrl: tailnet.NewBasicTelemetryController(logger),
}
}
@ -124,9 +114,6 @@ func (tac *tailnetAPIConnector) runConnector(conn tailnetConn) {
if err != nil {
continue
}
tac.clientMu.Lock()
tac.client = tailnetClient
tac.clientMu.Unlock()
tac.logger.Debug(tac.ctx, "obtained tailnet API v2+ client")
tac.runConnectorOnce(tailnetClient)
tac.logger.Debug(tac.ctx, "tailnet API v2+ connection lost")
@ -141,9 +128,6 @@ var permanentErrorStatuses = []int{
}
func (tac *tailnetAPIConnector) dial() (proto.DRPCTailnetClient, error) {
if tac.customDialFn != nil {
return tac.customDialFn()
}
tac.logger.Debug(tac.ctx, "dialing Coder tailnet v2+ API")
u, err := url.Parse(tac.coordinateURL)
@ -228,6 +212,8 @@ func (tac *tailnetAPIConnector) runConnectorOnce(client proto.DRPCTailnetClient)
}
}()
tac.telCtrl.New(client) // synchronous, doesn't need a goroutine
refreshTokenCtx, refreshTokenCancel := context.WithCancel(tac.ctx)
wg := sync.WaitGroup{}
wg.Add(3)
@ -245,10 +231,7 @@ func (tac *tailnetAPIConnector) runConnectorOnce(client proto.DRPCTailnetClient)
// we do NOT want to gracefully disconnect on the coordinate() routine. So, we'll just
// close the underlying connection. This will trigger a retry of the control plane in
// run().
tac.clientMu.Lock()
client.DRPCConn().Close()
tac.client = nil
tac.clientMu.Unlock()
// Note that derpMap() logs it own errors, we don't bother here.
}
}()
@ -351,20 +334,5 @@ func (tac *tailnetAPIConnector) refreshToken(ctx context.Context, client proto.D
}
func (tac *tailnetAPIConnector) SendTelemetryEvent(event *proto.TelemetryEvent) {
tac.clientMu.RLock()
// We hold the lock for the entire telemetry request, but this would only block
// a coordinate retry, and closing the connection.
defer tac.clientMu.RUnlock()
if tac.client == nil || tac.telemetryUnavailable.Load() {
return
}
ctx, cancel := context.WithTimeout(tac.ctx, 5*time.Second)
defer cancel()
_, err := tac.client.PostTelemetry(ctx, &proto.TelemetryRequest{
Events: []*proto.TelemetryEvent{event},
})
if drpcerr.Code(err) == drpcerr.Unimplemented || drpc.ProtocolError.Has(err) && strings.Contains(err.Error(), "unknown rpc: ") {
tac.logger.Debug(tac.ctx, "attempted to send telemetry to a server that doesn't support it", slog.Error(err))
tac.telemetryUnavailable.Store(true)
}
tac.telCtrl.SendTelemetryEvent(event)
}