ooni-probe-cli/pkg/oonimkall/taskrunner.go
Simone Basso 9b08dcac3f
fix(oonimkall): only set annotations on success (#821)
This bug is one of these bugs that definitely help one to stay
humble and focused on improving the codebase.

Of course I `<facepalmed>` when I understood the root cause.

We did not move the annotations below the `if` which is checking
whether the measurement was successful when we refactored the
codebase to support returning multiple measurements per run, which
happened in https://github.com/ooni/probe-cli/pull/527.

While I am not going to whip myself too much because of this, it's
clearly a bummer that we didn't notice this bug back then. On top
of this, it's also quite sad it took us so much time to notice that
there was this bug inside the tree.

The lesson (hopefully) learned is probably that we need to be more
careful when we refactor and we should always ask the question of
whether, not only we have tests, but whether these tests could maybe
be improved to give us even more confidence about correctness.

The reference issue is https://github.com/ooni/probe/issues/2173.
2022-07-01 09:54:35 +02:00

351 lines
11 KiB
Go

package oonimkall
import (
"context"
"encoding/json"
"fmt"
"net/url"
"time"
"github.com/ooni/probe-cli/v3/internal/engine"
"github.com/ooni/probe-cli/v3/internal/model"
"github.com/ooni/probe-cli/v3/internal/runtimex"
)
// runnerForTask runs a specific task
type runnerForTask struct {
emitter *taskEmitterWrapper
kvStoreBuilder taskKVStoreFSBuilder
sessionBuilder taskSessionBuilder
settings *settings
}
var _ taskRunner = &runnerForTask{}
// newRunner creates a new task runner
func newRunner(settings *settings, emitter taskEmitter) *runnerForTask {
return &runnerForTask{
emitter: &taskEmitterWrapper{emitter},
kvStoreBuilder: &taskKVStoreFSBuilderEngine{},
sessionBuilder: &taskSessionBuilderEngine{},
settings: settings,
}
}
// failureInvalidVersion is the failure returned when Version is invalid
const failureInvalidVersion = "invalid Settings.Version number"
func (r *runnerForTask) hasUnsupportedSettings() bool {
if r.settings.Version < taskABIVersion {
r.emitter.EmitFailureStartup(failureInvalidVersion)
return true
}
return false
}
func (r *runnerForTask) newsession(ctx context.Context, logger model.Logger) (taskSession, error) {
kvstore, err := r.kvStoreBuilder.NewFS(r.settings.StateDir)
if err != nil {
return nil, err
}
var proxyURL *url.URL
if r.settings.Proxy != "" {
var err error
proxyURL, err = url.Parse(r.settings.Proxy)
if err != nil {
return nil, err
}
}
config := engine.SessionConfig{
KVStore: kvstore,
Logger: logger,
ProxyURL: proxyURL,
SoftwareName: r.settings.Options.SoftwareName,
SoftwareVersion: r.settings.Options.SoftwareVersion,
TempDir: r.settings.TempDir,
TunnelDir: r.settings.TunnelDir,
}
if r.settings.Options.ProbeServicesBaseURL != "" {
config.AvailableProbeServices = []model.OOAPIService{{
Type: "https",
Address: r.settings.Options.ProbeServicesBaseURL,
}}
}
return r.sessionBuilder.NewSession(ctx, config)
}
// contextForExperiment ensurs that for measuring we only use an
// interruptible context when we can interrupt the experiment
func (r *runnerForTask) contextForExperiment(
ctx context.Context, builder taskExperimentBuilder,
) context.Context {
if builder.Interruptible() {
return ctx
}
return context.Background()
}
type runnerCallbacks struct {
emitter taskEmitter
}
func (cb *runnerCallbacks) OnProgress(percentage float64, message string) {
cb.emitter.Emit(eventTypeStatusProgress, eventStatusProgress{
Percentage: 0.4 + (percentage * 0.6), // open report is 40%
Message: message,
})
}
// Run runs the runner until completion. The context argument controls
// when to stop when processing multiple inputs, as well as when to stop
// experiments explicitly marked as interruptible.
func (r *runnerForTask) Run(rootCtx context.Context) {
// Implementation note: this function uses these contexts:
//
// - rootCtx is the root context and is controlled by the user;
//
// - measCtx derives from rootCtx and is possibly tied to the
// maximum runtime and is used to choose when to stop measuring;
//
// - submitCtx is like measCtx but, in case we're using a max
// runtime, is given more time to finish submitting.
//
// See https://github.com/ooni/probe/issues/2037.
var logger model.Logger = newTaskLogger(r.emitter, r.settings.LogLevel)
r.emitter.Emit(eventTypeStatusQueued, eventEmpty{})
if r.hasUnsupportedSettings() {
// event failureStartup already emitted
return
}
r.emitter.Emit(eventTypeStatusStarted, eventEmpty{})
sess, err := r.newsession(rootCtx, logger)
if err != nil {
r.emitter.EmitFailureStartup(err.Error())
return
}
endEvent := new(eventStatusEnd)
defer func() {
sess.Close()
r.emitter.Emit(eventTypeStatusEnd, endEvent)
}()
builder, err := sess.NewExperimentBuilderByName(r.settings.Name)
if err != nil {
r.emitter.EmitFailureStartup(err.Error())
return
}
logger.Info("Looking up OONI backends... please, be patient")
if err := sess.MaybeLookupBackendsContext(rootCtx); err != nil {
r.emitter.EmitFailureStartup(err.Error())
return
}
r.emitter.EmitStatusProgress(0.1, "contacted bouncer")
logger.Info("Looking up your location... please, be patient")
if err := sess.MaybeLookupLocationContext(rootCtx); err != nil {
r.emitter.EmitFailureGeneric(eventTypeFailureIPLookup, err.Error())
r.emitter.EmitFailureGeneric(eventTypeFailureASNLookup, err.Error())
r.emitter.EmitFailureGeneric(eventTypeFailureCCLookup, err.Error())
r.emitter.EmitFailureGeneric(eventTypeFailureResolverLookup, err.Error())
return
}
r.emitter.EmitStatusProgress(0.2, "geoip lookup")
r.emitter.EmitStatusProgress(0.3, "resolver lookup")
r.emitter.Emit(eventTypeStatusGeoIPLookup, eventStatusGeoIPLookup{
ProbeIP: sess.ProbeIP(),
ProbeASN: sess.ProbeASNString(),
ProbeCC: sess.ProbeCC(),
ProbeNetworkName: sess.ProbeNetworkName(),
})
r.emitter.Emit(eventTypeStatusResolverLookup, eventStatusResolverLookup{
ResolverASN: sess.ResolverASNString(),
ResolverIP: sess.ResolverIP(),
ResolverNetworkName: sess.ResolverNetworkName(),
})
builder.SetCallbacks(&runnerCallbacks{emitter: r.emitter})
// TODO(bassosimone): replace the following code with an
// invocation of the InputLoader. Since I am making these
// changes before a release and I've already changed the
// code a lot, I'd rather avoid changing it even more,
// for the following reason:
//
// If we add an call InputLoader here, this code will
// magically invoke check-in for InputOrQueryBackend,
// which we need to make sure the app can handle. This is
// the main reason why now I don't fill like properly
// fixing this code and use InputLoader: too much work
// in too little time, so mistakes more likely.
//
// In fact, our current app assumes that it's its
// responsibility to load the inputs, not oonimkall's.
switch builder.InputPolicy() {
case engine.InputOrQueryBackend, engine.InputStrictlyRequired:
if len(r.settings.Inputs) <= 0 {
r.emitter.EmitFailureStartup("no input provided")
return
}
case engine.InputOrStaticDefault:
if len(r.settings.Inputs) <= 0 {
inputs, err := engine.StaticBareInputForExperiment(r.settings.Name)
if err != nil {
r.emitter.EmitFailureStartup("no default static input for this experiment")
return
}
r.settings.Inputs = inputs
}
case engine.InputOptional:
if len(r.settings.Inputs) <= 0 {
r.settings.Inputs = append(r.settings.Inputs, "")
}
default: // treat this case as engine.InputNone.
if len(r.settings.Inputs) > 0 {
r.emitter.EmitFailureStartup("experiment does not accept input")
return
}
r.settings.Inputs = append(r.settings.Inputs, "")
}
experiment := builder.NewExperimentInstance()
defer func() {
endEvent.DownloadedKB = experiment.KibiBytesReceived()
endEvent.UploadedKB = experiment.KibiBytesSent()
}()
if !r.settings.Options.NoCollector {
logger.Info("Opening report... please, be patient")
if err := experiment.OpenReportContext(rootCtx); err != nil {
r.emitter.EmitFailureGeneric(eventTypeFailureReportCreate, err.Error())
return
}
r.emitter.EmitStatusProgress(0.4, "open report")
r.emitter.Emit(eventTypeStatusReportCreate, eventStatusReportGeneric{
ReportID: experiment.ReportID(),
})
}
measCtx, measCancel := context.WithCancel(rootCtx)
defer measCancel()
submitCtx, submitCancel := context.WithCancel(rootCtx)
defer submitCancel()
// This deviates a little bit from measurement-kit, for which
// a zero timeout is actually valid. Since it does not make much
// sense, here we're changing the behaviour.
//
// See https://github.com/measurement-kit/measurement-kit/issues/1922
if r.settings.Options.MaxRuntime > 0 {
// We want to honour max_runtime only when we're running an
// experiment that clearly wants specific input. We could refine
// this policy in the future, but for now this covers in a
// reasonable way web connectivity, so we should be ok.
switch builder.InputPolicy() {
case engine.InputOrQueryBackend, engine.InputStrictlyRequired:
var (
cancelMeas context.CancelFunc
cancelSubmit context.CancelFunc
)
// We give the context used for submitting extra time so that
// it's possible to submit the last measurement.
//
// See https://github.com/ooni/probe/issues/2037 for more info.
maxRuntime := time.Duration(r.settings.Options.MaxRuntime) * time.Second
measCtx, cancelMeas = context.WithTimeout(measCtx, maxRuntime)
defer cancelMeas()
maxRuntime += 30 * time.Second
submitCtx, cancelSubmit = context.WithTimeout(submitCtx, maxRuntime)
defer cancelSubmit()
}
}
inputCount := len(r.settings.Inputs)
start := time.Now()
inflatedMaxRuntime := r.settings.Options.MaxRuntime + r.settings.Options.MaxRuntime/10
eta := start.Add(time.Duration(inflatedMaxRuntime) * time.Second)
for idx, input := range r.settings.Inputs {
if measCtx.Err() != nil {
break
}
logger.Infof("Starting measurement with index %d", idx)
r.emitter.Emit(eventTypeStatusMeasurementStart, eventMeasurementGeneric{
Idx: int64(idx),
Input: input,
})
if input != "" && inputCount > 0 {
var percentage float64
if r.settings.Options.MaxRuntime > 0 {
now := time.Now()
percentage = (now.Sub(start).Seconds()/eta.Sub(start).Seconds())*0.6 + 0.4
} else {
percentage = (float64(idx)/float64(inputCount))*0.6 + 0.4
}
r.emitter.EmitStatusProgress(percentage, fmt.Sprintf(
"processing %s", input,
))
}
m, err := experiment.MeasureWithContext(
r.contextForExperiment(measCtx, builder),
input,
)
if builder.Interruptible() && measCtx.Err() != nil {
// We want to stop here only if interruptible otherwise we want to
// submit measurement and stop at beginning of next iteration
break
}
if err != nil {
r.emitter.Emit(eventTypeFailureMeasurement, eventMeasurementGeneric{
Failure: err.Error(),
Idx: int64(idx),
Input: input,
})
// Historical note: here we used to fallthrough but, since we have
// implemented async measurements, the case where there is an error
// and we also have a valid measurement cant't happen anymore. So,
// now the only valid strategy here is to continue.
continue
}
m.AddAnnotations(r.settings.Annotations)
data, err := json.Marshal(m)
runtimex.PanicOnError(err, "measurement.MarshalJSON failed")
r.emitter.Emit(eventTypeMeasurement, eventMeasurementGeneric{
Idx: int64(idx),
Input: input,
JSONStr: string(data),
})
if !r.settings.Options.NoCollector {
logger.Info("Submitting measurement... please, be patient")
err := experiment.SubmitAndUpdateMeasurementContext(submitCtx, m)
warnOnFailure(logger, "cannot submit measurement", err)
r.emitter.Emit(measurementSubmissionEventName(err), eventMeasurementGeneric{
Idx: int64(idx),
Input: input,
JSONStr: string(data),
Failure: measurementSubmissionFailure(err),
})
}
r.emitter.Emit(eventTypeStatusMeasurementDone, eventMeasurementGeneric{
Idx: int64(idx),
Input: input,
})
}
}
func warnOnFailure(logger model.Logger, message string, err error) {
if err != nil {
logger.Warnf("%s: %s (%+v)", message, err.Error(), err)
}
}
func measurementSubmissionEventName(err error) string {
if err != nil {
return eventTypeFailureMeasurementSubmission
}
return eventTypeStatusMeasurementSubmission
}
func measurementSubmissionFailure(err error) string {
if err != nil {
return err.Error()
}
return ""
}