Files
coder/coderd/agentapi/stats.go
T
Callum Styan 8ed1c1d372 perf: reduce calls to GetWorkspaceByAgentID in GetWorkspaceAgentByID (#21046)
This PR piggy backs on the agent API cached workspace added in an earlier PR to provide a fast path for avoiding `GetWorkspaceByAgentID` calls in dbauthz's `GetWorkspaceAgentByID`. This query is not the most expensive, but has a significant call volume at ~16 million calls per week.

Signed-off-by: Callum Styan <callumstyan@gmail.com>
2025-12-10 14:03:24 -08:00

107 lines
3.0 KiB
Go

package agentapi
import (
"context"
"time"
"golang.org/x/xerrors"
"google.golang.org/protobuf/types/known/durationpb"
"cdr.dev/slog"
agentproto "github.com/coder/coder/v2/agent/proto"
"github.com/coder/coder/v2/coderd/database"
"github.com/coder/coder/v2/coderd/database/dbauthz"
"github.com/coder/coder/v2/coderd/database/dbtime"
"github.com/coder/coder/v2/coderd/workspacestats"
"github.com/coder/coder/v2/codersdk"
)
type StatsAPI struct {
AgentFn func(context.Context) (database.WorkspaceAgent, error)
Workspace *CachedWorkspaceFields
Database database.Store
Log slog.Logger
StatsReporter *workspacestats.Reporter
AgentStatsRefreshInterval time.Duration
Experiments codersdk.Experiments
TimeNowFn func() time.Time // defaults to dbtime.Now()
}
func (a *StatsAPI) now() time.Time {
if a.TimeNowFn != nil {
return a.TimeNowFn()
}
return dbtime.Now()
}
func (a *StatsAPI) UpdateStats(ctx context.Context, req *agentproto.UpdateStatsRequest) (*agentproto.UpdateStatsResponse, error) {
res := &agentproto.UpdateStatsResponse{
ReportInterval: durationpb.New(a.AgentStatsRefreshInterval),
}
// An empty stat means it's just looking for the report interval.
if req.Stats == nil {
return res, nil
}
// Inject RBAC object into context for dbauthz fast path, avoid having to
// call GetWorkspaceAgentByID on every stats update.
rbacCtx := ctx
if dbws, ok := a.Workspace.AsWorkspaceIdentity(); ok {
var err error
rbacCtx, err = dbauthz.WithWorkspaceRBAC(ctx, dbws.RBACObject())
if err != nil {
// Don't error level log here, will exit the function. We want to fall back to GetWorkspaceByAgentID.
//nolint:gocritic
a.Log.Debug(ctx, "Cached workspace was present but RBAC object was invalid", slog.F("err", err))
}
}
workspaceAgent, err := a.AgentFn(rbacCtx)
if err != nil {
return nil, err
}
// If cache is empty (prebuild or invalid), fall back to DB
var ws database.WorkspaceIdentity
var ok bool
if ws, ok = a.Workspace.AsWorkspaceIdentity(); !ok {
w, err := a.Database.GetWorkspaceByAgentID(ctx, workspaceAgent.ID)
if err != nil {
return nil, xerrors.Errorf("get workspace by agent ID %q: %w", workspaceAgent.ID, err)
}
ws = database.WorkspaceIdentityFromWorkspace(w)
}
a.Log.Debug(ctx, "read stats report",
slog.F("interval", a.AgentStatsRefreshInterval),
slog.F("workspace_id", ws.ID),
slog.F("payload", req),
)
if a.Experiments.Enabled(codersdk.ExperimentWorkspaceUsage) {
// while the experiment is enabled we will not report
// session stats from the agent. This is because it is
// being handled by the CLI and the postWorkspaceUsage route.
req.Stats.SessionCountSsh = 0
req.Stats.SessionCountJetbrains = 0
req.Stats.SessionCountVscode = 0
req.Stats.SessionCountReconnectingPty = 0
}
err = a.StatsReporter.ReportAgentStats(
ctx,
a.now(),
ws,
workspaceAgent,
req.Stats,
false,
)
if err != nil {
return nil, xerrors.Errorf("report agent stats: %w", err)
}
return res, nil
}