Files
releases/runtime/runtime.go
T
Stephan Renatus 516dd47dd1 runtime+storage: integrate disk storage
With this change, the disk backend (badger) becomes available for
use with the OPA runtime properly:

It can be configured using the `storage.disk` key in OPA's config
(see included documentation).

When enabled,
- any data or policies stored with OPA will persist over restarts
- per-query metrics related to disk usage are reported
- Prometheus metrics per storage operation are exported

The main intention behind this feature is to optimize memory usage:
OPA can now operate on more data than fits into the allotted memory
resources. It is NOT meant to be used as a primary source of truth:
there are no backup/restore or desaster recovery procedures -- you
MUST secure the means to restore the data stored with OPA's disk
storage by yourself.

See also #4014. Future improvements around bundle loading are
planned.

Some notes on details:

storage/disk: impose same locking regime used with inmem

With this setup, we'll ensure:

- there is only one open write txn at a time
- there are any number of open read txns at a time
- writes are blocked when reads are inflight
- during a commit (and triggers being run), no read txns can be created

This is to ensure the same atomic policy update semantics when using
'disk" as we have with "inmem". We're basically opting out of badger's
currency control and transactionality guarantees. This is because we
cannot piggy back on that to ensure the atomic update we want.

There might be other ways -- using subscribers, and blocking in some
other place -- but this one seems preferrable since it mirrors inmem.

Part of the problem is ErrTxnTooLarge, and committing and renewing
txns when it occurs: that, which is the prescribed solution to txns
growing too big, also means that reads can see half of the "logical"
transaction having been committed, while the rest is still getting
processed.

Another approach would have been using `WriteBatch`, but that won't
let us read from the batch, only apply Set and Delete operations.
We currently need to read (via an iterator) to figure out if we
need to delete keys to replace something in the store.  There is
no DropPrefix operation on the badger txn, or the WriteBatch API.

storage/disk: remove commit-and-renew-txn code for txn-too-big errors

This would break transactional guarantees we care about: while there
can be only one write transaction at a time, read transactions may
happen while a write txn is underway -- with this commit-and-reset
logic, those would read partial data.

Now, the error will be returned to the caller. The maximum txn size
depends on the size of memtables, and could be tweaked manually.
In general, the caller should try to push multiple smaller increments
of the data.

storage/disk: implement noop MakeDir

The MakeDir operation as implemented in the backend-agnostic storage
code has become an issue with the disk store: to write /foo/bar/baz,
we'd have to read /foo (among other subdirs), and that can be _much_
work for the disk backend. With inmem, it's cheap, so this wasn't
problematic before.

Some of the storage/disk/txn.go logic had to be adjusted to properly
do the MakeDir steps implicitly.

The index argument addition to patch() in storage/disk/txn.go was
necessary to keep the error messages conforming to the previous
code path: previously, conflicts (arrays indexed as objects) would
be surfaced in the MakeDir step, now it's entangled with the patch
calculation.

storage/disk: check ctx.Err() in List/Get operations

This won't abort reading a single key, but it will abort iterations.

storage/disk: support patterns in partitions

There is a potential clash here: "*", the path wildcard, is
a valid path section. However, it only affects the case when
a user would want to have a partition at

    /foo/*/bar

and would really mean "*", and not the wildcard.

Storing data at /foo/*/bar with a literal "*" won't be treated
differently than storing something at /fo/xyz/bar.

storage/disk: keep per-txn-type histograms of stats

This is done by reading off the metrics on commit, and shovelling
their numbers into the prometheus collector.

NOTE: if you were to share a metrics object among multiple transactions,
the results would be skewed, as it's not reset. However, our server
handlers don't do that.

storage/disk: opt out of badger's conflict detection

With only one write transaction in flight at any time, the situation
that badger guards against cannot happen:

A transaction has written to a key after the current, to-be-committed
transaction has last read that key from the store.

Since it can't happen, we can ignore the bookkeeping involved. This
improves the time it takes to overwrite existing keys.

Signed-off-by: Stephan Renatus <stephan.renatus@gmail.com>
2022-03-30 10:25:45 +02:00

897 lines
27 KiB
Go

// Copyright 2016 The OPA Authors. All rights reserved.
// Use of this source code is governed by an Apache2
// license that can be found in the LICENSE file.
package runtime
import (
"bytes"
"context"
"crypto/rand"
"crypto/tls"
"crypto/x509"
"fmt"
"io"
mr "math/rand"
"os"
"os/signal"
"os/user"
"strings"
"sync"
"syscall"
"time"
"github.com/fsnotify/fsnotify"
"github.com/gorilla/mux"
"go.opentelemetry.io/otel/exporters/otlp/otlptrace"
"go.uber.org/automaxprocs/maxprocs"
"github.com/open-policy-agent/opa/ast"
"github.com/open-policy-agent/opa/bundle"
"github.com/open-policy-agent/opa/internal/config"
internal_tracing "github.com/open-policy-agent/opa/internal/distributedtracing"
internal_logging "github.com/open-policy-agent/opa/internal/logging"
"github.com/open-policy-agent/opa/internal/prometheus"
"github.com/open-policy-agent/opa/internal/report"
"github.com/open-policy-agent/opa/internal/runtime"
initload "github.com/open-policy-agent/opa/internal/runtime/init"
"github.com/open-policy-agent/opa/internal/uuid"
"github.com/open-policy-agent/opa/loader"
"github.com/open-policy-agent/opa/logging"
"github.com/open-policy-agent/opa/metrics"
"github.com/open-policy-agent/opa/plugins"
"github.com/open-policy-agent/opa/plugins/discovery"
"github.com/open-policy-agent/opa/plugins/logs"
"github.com/open-policy-agent/opa/repl"
"github.com/open-policy-agent/opa/server"
"github.com/open-policy-agent/opa/storage"
"github.com/open-policy-agent/opa/storage/disk"
"github.com/open-policy-agent/opa/storage/inmem"
"github.com/open-policy-agent/opa/tracing"
"github.com/open-policy-agent/opa/util"
"github.com/open-policy-agent/opa/version"
)
var (
registeredPlugins map[string]plugins.Factory
registeredPluginsMux sync.Mutex
)
const (
// default interval between OPA version report uploads
defaultUploadIntervalSec = int64(3600)
)
// RegisterPlugin registers a plugin factory with the runtime
// package. When the runtime is created, the factories are used to parse
// plugin configuration and instantiate plugins. If no configuration is
// provided, plugins are not instantiated. This function is idempotent.
func RegisterPlugin(name string, factory plugins.Factory) {
registeredPluginsMux.Lock()
defer registeredPluginsMux.Unlock()
registeredPlugins[name] = factory
}
// Params stores the configuration for an OPA instance.
type Params struct {
// Globally unique identifier for this OPA instance. If an ID is not specified,
// the runtime will generate one.
ID string
// Addrs are the listening addresses that the OPA server will bind to.
Addrs *[]string
// DiagnosticAddrs are the listening addresses that the OPA server will bind to
// for read-only diagnostic API's (/health, /metrics, etc)
DiagnosticAddrs *[]string
// H2CEnabled flag controls whether OPA will allow H2C (HTTP/2 cleartext) on
// HTTP listeners.
H2CEnabled bool
// Authentication is the type of authentication scheme to use.
Authentication server.AuthenticationScheme
// Authorization is the type of authorization scheme to use.
Authorization server.AuthorizationScheme
// Certificate is the certificate to use in server-mode. If the certificate
// is nil, the server will NOT use TLS.
Certificate *tls.Certificate
// CertificateFile and CertificateKeyFile are the paths to the cert and its
// keyfile. It'll be used to periodically reload the files from disk if they
// have changed. The server will attempt to refresh every 5 minutes, unless
// a different CertificateRefresh time.Duration is provided
CertificateFile string
CertificateKeyFile string
CertificateRefresh time.Duration
// CertPool holds the CA certs trusted by the OPA server.
CertPool *x509.CertPool
// MinVersion contains the minimum TLS version that is acceptable.
// If zero, TLS 1.2 is currently taken as the minimum.
MinTLSVersion uint16
// HistoryPath is the filename to store the interactive shell user
// input history.
HistoryPath string
// Output format controls how the REPL will print query results.
// Default: "pretty".
OutputFormat string
// Paths contains filenames of base documents and policy modules to load on
// startup. Data files may be prefixed with "<dotted-path>:" to indicate
// where the contained document should be loaded.
Paths []string
// Optional filter that will be passed to the file loader.
Filter loader.Filter
// BundleMode will enable treating the Paths provided as bundles rather than
// loading all data & policy files.
BundleMode bool
// Watch flag controls whether OPA will watch the Paths files for changes.
// If this flag is true, OPA will watch the Paths files for changes and
// reload the storage layer each time they change. This is useful for
// interactive development.
Watch bool
// ErrorLimit is the number of errors the compiler will allow to occur before
// exiting early.
ErrorLimit int
// PprofEnabled flag controls whether pprof endpoints are enabled
PprofEnabled bool
// DecisionIDFactory generates decision IDs to include in API responses
// sent by the server (in response to Data API queries.)
DecisionIDFactory func() string
// Logging configures the logging behaviour.
Logging LoggingConfig
// Logger sets the logger implementation to use for debug logs.
Logger logging.Logger
// ConsoleLogger sets the logger implementation to use for console logs.
ConsoleLogger logging.Logger
// ConfigFile refers to the OPA configuration to load on startup.
ConfigFile string
// ConfigOverrides are overrides for the OPA configuration that are applied
// over top the config file They are in a list of key=value syntax that
// conform to the syntax defined in the `strval` package
ConfigOverrides []string
// ConfigOverrideFiles Similar to `ConfigOverrides` except they are in the
// form of `key=path/to/file`where the file contains the value to be used.
ConfigOverrideFiles []string
// Output is the output stream used when run as an interactive shell. This
// is mostly for test purposes.
Output io.Writer
// GracefulShutdownPeriod is the time (in seconds) to wait for the http
// server to shutdown gracefully.
GracefulShutdownPeriod int
// ShutdownWaitPeriod is the time (in seconds) to wait before initiating shutdown.
ShutdownWaitPeriod int
// EnableVersionCheck flag controls whether OPA will report its version to an external service.
// If this flag is true, OPA will report its version to the external service
EnableVersionCheck bool
// BundleVerificationConfig sets the key configuration used to verify a signed bundle
BundleVerificationConfig *bundle.VerificationConfig
// SkipBundleVerification flag controls whether OPA will verify a signed bundle
SkipBundleVerification bool
// ReadyTimeout flag controls if and for how long OPA server will wait (in seconds) for
// configured bundles and plugins to be activated/ready before listening for traffic.
// A value of 0 or less means no wait is exercised.
ReadyTimeout int
// Router is the router to which handlers for the REST API are added.
// Router uses a first-matching-route-wins strategy, so no existing routes are overridden
// If it is nil, a new mux.Router will be created
Router *mux.Router
// DiskStorage, if set, will make the runtime instantiate a disk-backed storage
// implementation (instead of the default, in-memory store).
// It can also be enabled via config, and this runtime field takes precedence.
DiskStorage *disk.Options
DistributedTracingOpts tracing.Options
}
// LoggingConfig stores the configuration for OPA's logging behaviour.
type LoggingConfig struct {
Level string
Format string
}
// NewParams returns a new Params object.
func NewParams() Params {
return Params{
Output: os.Stdout,
BundleMode: false,
EnableVersionCheck: false,
}
}
// Runtime represents a single OPA instance.
type Runtime struct {
Params Params
Store storage.Store
Manager *plugins.Manager
logger logging.Logger
server *server.Server
metrics *prometheus.Provider
reporter *report.Reporter
traceExporter *otlptrace.Exporter
serverInitialized bool
serverInitMtx sync.RWMutex
done chan struct{}
}
// NewRuntime returns a new Runtime object initialized with params. Clients must
// call StartServer() or StartREPL() to start the runtime in either mode.
func NewRuntime(ctx context.Context, params Params) (*Runtime, error) {
if params.ID == "" {
var err error
params.ID, err = generateInstanceID()
if err != nil {
return nil, err
}
}
level, err := internal_logging.GetLevel(params.Logging.Level)
if err != nil {
return nil, err
}
// NOTE(tsandall): This is a temporary hack to ensure that log formatting
// and leveling is applied correctly. Currently there are a few places where
// the global logger is used as a fallback, however, that fallback _should_
// never be used. This ensures that _if_ the fallback is used accidentally,
// that the logging configuration is applied. Once we remove all usage of
// the global logger and we remove the API that allows callers to access the
// global logger, we can remove this.
logging.Get().SetFormatter(internal_logging.GetFormatter(params.Logging.Format))
logging.Get().SetLevel(level)
var logger logging.Logger
if params.Logger != nil {
logger = params.Logger
} else {
stdLogger := logging.New()
stdLogger.SetLevel(level)
stdLogger.SetFormatter(internal_logging.GetFormatter(params.Logging.Format))
logger = stdLogger
}
config, err := config.Load(params.ConfigFile, params.ConfigOverrides, params.ConfigOverrideFiles)
if err != nil {
return nil, fmt.Errorf("config error: %w", err)
}
var reporter *report.Reporter
if params.EnableVersionCheck {
var err error
reporter, err = report.New(params.ID, report.Options{Logger: logger})
if err != nil {
return nil, fmt.Errorf("config error: %w", err)
}
}
loaded, err := initload.LoadPaths(params.Paths, params.Filter, params.BundleMode, params.BundleVerificationConfig, params.SkipBundleVerification)
if err != nil {
return nil, fmt.Errorf("load error: %w", err)
}
info, err := runtime.Term(runtime.Params{Config: config})
if err != nil {
return nil, err
}
consoleLogger := params.ConsoleLogger
if consoleLogger == nil {
l := logging.New()
l.SetFormatter(internal_logging.GetFormatter(params.Logging.Format))
consoleLogger = l
}
if params.Router == nil {
params.Router = mux.NewRouter()
}
metrics := prometheus.New(metrics.New(), errorLogger(logger))
var store storage.Store
if params.DiskStorage == nil {
params.DiskStorage, err = disk.OptionsFromConfig(config, params.ID)
if err != nil {
return nil, fmt.Errorf("parse disk store configuration: %w", err)
}
}
if params.DiskStorage != nil {
store, err = disk.New(ctx, logger, metrics, *params.DiskStorage)
if err != nil {
return nil, fmt.Errorf("initialize disk store: %w", err)
}
} else {
store = inmem.New()
}
manager, err := plugins.New(config,
params.ID,
store,
plugins.Info(info),
plugins.InitBundles(loaded.Bundles),
plugins.InitFiles(loaded.Files),
plugins.MaxErrors(params.ErrorLimit),
plugins.GracefulShutdownPeriod(params.GracefulShutdownPeriod),
plugins.ConsoleLogger(consoleLogger),
plugins.Logger(logger),
plugins.EnablePrintStatements(logger.GetLevel() >= logging.Info),
plugins.PrintHook(loggingPrintHook{logger: logger}),
plugins.WithRouter(params.Router),
plugins.WithPrometheusRegister(metrics))
if err != nil {
return nil, fmt.Errorf("config error: %w", err)
}
if err := manager.Init(ctx); err != nil {
return nil, fmt.Errorf("initialization error: %w", err)
}
traceExporter, distributedTracingOpts, err := internal_tracing.Init(ctx, config, params.ID)
if err != nil {
return nil, fmt.Errorf("config error: %w", err)
}
if distributedTracingOpts != nil {
params.DistributedTracingOpts = distributedTracingOpts
}
disco, err := discovery.New(manager, discovery.Factories(registeredPlugins), discovery.Metrics(metrics))
if err != nil {
return nil, fmt.Errorf("config error: %w", err)
}
manager.Register(discovery.Name, disco)
rt := &Runtime{
Store: manager.Store,
Params: params,
Manager: manager,
logger: logger,
metrics: metrics,
reporter: reporter,
serverInitialized: false,
traceExporter: traceExporter,
}
return rt, nil
}
// StartServer starts the runtime in server mode. This function will block the
// calling goroutine and will exit the program on error.
func (rt *Runtime) StartServer(ctx context.Context) {
err := rt.Serve(ctx)
if err != nil {
os.Exit(1)
}
}
// Serve will start a new REST API server and listen for requests. This
// will block until either: an error occurs, the context is canceled, or
// a SIGTERM or SIGKILL signal is sent.
func (rt *Runtime) Serve(ctx context.Context) error {
if rt.Params.Addrs == nil {
return fmt.Errorf("at least one address must be configured in runtime parameters")
}
if rt.Params.DiagnosticAddrs == nil {
rt.Params.DiagnosticAddrs = &[]string{}
}
rt.logger.WithFields(map[string]interface{}{
"addrs": *rt.Params.Addrs,
"diagnostic-addrs": *rt.Params.DiagnosticAddrs,
}).Info("Initializing server.")
if rt.Params.Authorization == server.AuthorizationOff && rt.Params.Authentication == server.AuthenticationToken {
rt.logger.Error("Token authentication enabled without authorization. Authentication will be ineffective. See https://www.openpolicyagent.org/docs/latest/security/#authentication-and-authorization for more information.")
}
usr, err := user.Current()
if err != nil {
rt.logger.Debug("Failed to determine uid/gid of process owner")
} else if usr.Uid == "0" || usr.Gid == "0" {
message := "OPA running with uid or gid 0. Running OPA with root privileges is not recommended."
if os.Getenv("OPA_DOCKER_IMAGE") == "official" {
message += " Use the -rootless image to avoid running with root privileges. This will be made the default in later OPA releases."
}
rt.logger.Warn(message)
}
// NOTE(tsandall): at some point, hopefully we can remove this because the
// Go runtime will just do the right thing. Until then, try to set
// GOMAXPROCS based on the CPU quota applied to the process.
undo, err := maxprocs.Set(maxprocs.Logger(func(f string, a ...interface{}) {
rt.logger.Debug(f, a...)
}))
if err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Debug("Failed to set GOMAXPROCS from CPU quota.")
}
defer undo()
if err := rt.Manager.Start(ctx); err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Error("Failed to start plugins.")
return err
}
defer rt.Manager.Stop(ctx)
if rt.traceExporter != nil {
if err := rt.traceExporter.Start(ctx); err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Error("Failed to start OpenTelemetry trace exporter.")
return err
}
defer func() {
err := rt.traceExporter.Shutdown(ctx)
if err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Error("Failed to shutdown OpenTelemetry trace exporter gracefully.")
}
}()
}
rt.server = server.New().
WithRouter(rt.Params.Router).
WithStore(rt.Store).
WithManager(rt.Manager).
WithCompilerErrorLimit(rt.Params.ErrorLimit).
WithPprofEnabled(rt.Params.PprofEnabled).
WithAddresses(*rt.Params.Addrs).
WithH2CEnabled(rt.Params.H2CEnabled).
WithCertificate(rt.Params.Certificate).
WithCertificatePaths(rt.Params.CertificateFile, rt.Params.CertificateKeyFile, rt.Params.CertificateRefresh).
WithCertPool(rt.Params.CertPool).
WithAuthentication(rt.Params.Authentication).
WithAuthorization(rt.Params.Authorization).
WithDecisionIDFactory(rt.decisionIDFactory).
WithDecisionLoggerWithErr(rt.decisionLogger).
WithRuntime(rt.Manager.Info).
WithMetrics(rt.metrics).
WithMinTLSVersion(rt.Params.MinTLSVersion).
WithDistributedTracingOpts(rt.Params.DistributedTracingOpts)
if rt.Params.DiagnosticAddrs != nil {
rt.server = rt.server.WithDiagnosticAddresses(*rt.Params.DiagnosticAddrs)
}
rt.server, err = rt.server.Init(ctx)
if err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Error("Unable to initialize server.")
return err
}
if rt.Params.Watch {
if err := rt.startWatcher(ctx, rt.Params.Paths, rt.onReloadLogger); err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Error("Unable to open watch.")
return err
}
}
if rt.Params.EnableVersionCheck {
d := time.Duration(int64(time.Second) * defaultUploadIntervalSec)
rt.done = make(chan struct{})
go rt.checkOPAUpdateLoop(ctx, d, rt.done)
}
defer func() {
if rt.done != nil {
rt.done <- struct{}{}
}
}()
rt.server.Handler = NewLoggingHandler(rt.logger, rt.server.Handler)
rt.server.DiagnosticHandler = NewLoggingHandler(rt.logger, rt.server.DiagnosticHandler)
if err := rt.waitPluginsReady(
100*time.Millisecond,
time.Second*time.Duration(rt.Params.ReadyTimeout)); err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Error("Failed to wait for plugins activation.")
return err
}
loops, err := rt.server.Listeners()
if err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Error("Unable to create listeners.")
return err
}
errc := make(chan error)
for _, loop := range loops {
go func(serverLoop func() error) {
errc <- serverLoop()
}(loop)
}
// Buffer one element as os/signal uses non-blocking channel sends.
// This prevents potentially dropping the first element and failing to shut
// down gracefully. A buffer of 1 is sufficient as we're just looking for a
// one-time shutdown signal.
signalc := make(chan os.Signal, 1)
signal.Notify(signalc, syscall.SIGINT, syscall.SIGTERM)
// Note that there is a small chance the socket of the server listener is still
// closed by the time this block is executed, due to the serverLoop above
// executing in a goroutine.
rt.serverInitMtx.Lock()
rt.serverInitialized = true
rt.serverInitMtx.Unlock()
rt.Manager.ServerInitialized()
rt.logger.Debug("Server initialized.")
for {
select {
case <-ctx.Done():
return rt.gracefulServerShutdown(rt.server)
case <-signalc:
return rt.gracefulServerShutdown(rt.server)
case err := <-errc:
rt.logger.WithFields(map[string]interface{}{"err": err}).Error("Listener failed.")
os.Exit(1)
}
}
}
// Addrs returns a list of addresses that the runtime is listening on (when
// in server mode). Returns an empty list if it hasn't started listening.
func (rt *Runtime) Addrs() []string {
rt.serverInitMtx.RLock()
defer rt.serverInitMtx.RUnlock()
if !rt.serverInitialized {
return nil
}
return rt.server.Addrs()
}
// DiagnosticAddrs returns a list of diagnostic addresses that the runtime is
// listening on (when in server mode). Returns an empty list if it hasn't
// started listening.
func (rt *Runtime) DiagnosticAddrs() []string {
if rt.server == nil {
return nil
}
return rt.server.DiagnosticAddrs()
}
// StartREPL starts the runtime in REPL mode. This function will block the calling goroutine.
func (rt *Runtime) StartREPL(ctx context.Context) {
if err := rt.Manager.Start(ctx); err != nil {
fmt.Fprintln(rt.Params.Output, "error starting plugins:", err)
os.Exit(1)
}
defer rt.Manager.Stop(ctx)
banner := rt.getBanner()
repl := repl.New(rt.Store, rt.Params.HistoryPath, rt.Params.Output, rt.Params.OutputFormat, rt.Params.ErrorLimit, banner).
WithRuntime(rt.Manager.Info)
if rt.Params.Watch {
if err := rt.startWatcher(ctx, rt.Params.Paths, onReloadPrinter(rt.Params.Output)); err != nil {
fmt.Fprintln(rt.Params.Output, "error opening watch:", err)
os.Exit(1)
}
}
if rt.Params.EnableVersionCheck {
go func() {
repl.SetOPAVersionReport(rt.checkOPAUpdate(ctx).Slice())
}()
}
repl.Loop(ctx)
}
// SetDistributedTracingLogging configures the distributed tracing's ErrorHandler,
// and logger instances.
func (rt *Runtime) SetDistributedTracingLogging() {
internal_tracing.SetupLogging(rt.logger)
}
func (rt *Runtime) checkOPAUpdate(ctx context.Context) *report.DataResponse {
resp, _ := rt.reporter.SendReport(ctx)
return resp
}
func (rt *Runtime) checkOPAUpdateLoop(ctx context.Context, uploadDuration time.Duration, done chan struct{}) {
ticker := time.NewTicker(uploadDuration)
mr.Seed(time.Now().UnixNano())
for {
resp, err := rt.reporter.SendReport(ctx)
if err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Debug("Unable to send OPA version report.")
} else {
if resp.Latest.OPAUpToDate {
rt.logger.WithFields(map[string]interface{}{
"current_version": version.Version,
}).Debug("OPA is up to date.")
} else {
rt.logger.WithFields(map[string]interface{}{
"download_opa": resp.Latest.Download,
"release_notes": resp.Latest.ReleaseNotes,
"current_version": version.Version,
"latest_version": strings.TrimPrefix(resp.Latest.LatestRelease, "v"),
}).Info("OPA is out of date.")
}
}
select {
case <-ticker.C:
ticker.Stop()
newInterval := mr.Int63n(defaultUploadIntervalSec) + defaultUploadIntervalSec
ticker = time.NewTicker(time.Duration(int64(time.Second) * newInterval))
case <-done:
ticker.Stop()
return
}
}
}
func (rt *Runtime) decisionIDFactory() string {
if rt.Params.DecisionIDFactory != nil {
return rt.Params.DecisionIDFactory()
}
if logs.Lookup(rt.Manager) != nil {
return generateDecisionID()
}
return ""
}
func (rt *Runtime) decisionLogger(ctx context.Context, event *server.Info) error {
plugin := logs.Lookup(rt.Manager)
if plugin == nil {
return nil
}
return plugin.Log(ctx, event)
}
func (rt *Runtime) startWatcher(ctx context.Context, paths []string, onReload func(time.Duration, error)) error {
watcher, err := rt.getWatcher(paths)
if err != nil {
return err
}
go rt.readWatcher(ctx, watcher, paths, onReload)
return nil
}
func (rt *Runtime) readWatcher(ctx context.Context, watcher *fsnotify.Watcher, paths []string, onReload func(time.Duration, error)) {
for evt := range watcher.Events {
removalMask := fsnotify.Remove | fsnotify.Rename
mask := fsnotify.Create | fsnotify.Write | removalMask
if (evt.Op & mask) != 0 {
rt.logger.WithFields(map[string]interface{}{
"event": evt.String(),
}).Debug("Registered file event.")
t0 := time.Now()
removed := ""
if (evt.Op & removalMask) != 0 {
removed = evt.Name
}
err := rt.processWatcherUpdate(ctx, paths, removed)
onReload(time.Since(t0), err)
}
}
}
func (rt *Runtime) processWatcherUpdate(ctx context.Context, paths []string, removed string) error {
loaded, err := initload.LoadPaths(paths, rt.Params.Filter, rt.Params.BundleMode, nil, true)
if err != nil {
return err
}
removed = loader.CleanPath(removed)
return storage.Txn(ctx, rt.Store, storage.WriteParams, func(txn storage.Transaction) error {
if !rt.Params.BundleMode {
ids, err := rt.Store.ListPolicies(ctx, txn)
if err != nil {
return err
}
for _, id := range ids {
if id == removed {
if err := rt.Store.DeletePolicy(ctx, txn, id); err != nil {
return err
}
} else if _, exists := loaded.Files.Modules[id]; !exists {
// This branch get hit in two cases.
// 1. Another piece of code has access to the store and inserts
// a policy out-of-band.
// 2. In between FS notification and loader.Filtered() call above, a
// policy is removed from disk.
bs, err := rt.Store.GetPolicy(ctx, txn, id)
if err != nil {
return err
}
module, err := ast.ParseModule(id, string(bs))
if err != nil {
return err
}
loaded.Files.Modules[id] = &loader.RegoFile{
Name: id,
Raw: bs,
Parsed: module,
}
}
}
}
_, err := initload.InsertAndCompile(ctx, initload.InsertAndCompileOptions{
Store: rt.Store,
Txn: txn,
Files: loaded.Files,
Bundles: loaded.Bundles,
MaxErrors: -1,
})
if err != nil {
return err
}
return nil
})
}
func (rt *Runtime) getBanner() string {
var buf bytes.Buffer
fmt.Fprintf(&buf, "OPA %v (commit %v, built at %v)\n", version.Version, version.Vcs, version.Timestamp)
fmt.Fprintf(&buf, "\n")
fmt.Fprintf(&buf, "Run 'help' to see a list of commands and check for updates.\n")
return buf.String()
}
func (rt *Runtime) gracefulServerShutdown(s *server.Server) error {
if rt.Params.ShutdownWaitPeriod > 0 {
rt.logger.Info("Waiting %vs before initiating shutdown...", rt.Params.ShutdownWaitPeriod)
time.Sleep(time.Duration(rt.Params.ShutdownWaitPeriod) * time.Second)
}
rt.logger.Info("Shutting down...")
ctx, cancel := context.WithTimeout(context.Background(), time.Duration(rt.Params.GracefulShutdownPeriod)*time.Second)
defer cancel()
err := s.Shutdown(ctx)
if err != nil {
rt.logger.WithFields(map[string]interface{}{"err": err}).Error("Failed to shutdown server gracefully.")
return err
}
rt.logger.Info("Server shutdown.")
return nil
}
func (rt *Runtime) waitPluginsReady(checkInterval, timeout time.Duration) error {
if timeout <= 0 {
return nil
}
// check readiness of all plugins
pluginsReady := func() bool {
for _, status := range rt.Manager.PluginStatus() {
if status != nil && status.State != plugins.StateOK {
return false
}
}
return true
}
rt.logger.Debug("Waiting for plugins activation (%v).", timeout)
return util.WaitFunc(pluginsReady, checkInterval, timeout)
}
func (rt *Runtime) onReloadLogger(d time.Duration, err error) {
rt.logger.WithFields(map[string]interface{}{
"duration": d,
"err": err,
}).Info("Processed file watch event.")
}
func (rt *Runtime) getWatcher(rootPaths []string) (*fsnotify.Watcher, error) {
watchPaths, err := getWatchPaths(rootPaths)
if err != nil {
return nil, err
}
watcher, err := fsnotify.NewWatcher()
if err != nil {
return nil, err
}
for _, path := range watchPaths {
rt.logger.WithFields(map[string]interface{}{"path": path}).Debug("watching path")
if err := watcher.Add(path); err != nil {
return nil, err
}
}
return watcher, nil
}
func errorLogger(logger logging.Logger) func(attrs map[string]interface{}, f string, a ...interface{}) {
return func(attrs map[string]interface{}, f string, a ...interface{}) {
logger.WithFields(map[string]interface{}(attrs)).Error(f, a...)
}
}
func getWatchPaths(rootPaths []string) ([]string, error) {
paths := []string{}
for _, path := range rootPaths {
_, path = loader.SplitPrefix(path)
result, err := loader.Paths(path, true)
if err != nil {
return nil, err
}
paths = append(paths, loader.Dirs(result)...)
}
return paths, nil
}
func onReloadPrinter(output io.Writer) func(time.Duration, error) {
return func(d time.Duration, err error) {
if err != nil {
fmt.Fprintf(output, "\n# reload error (took %v): %v", d, err)
} else {
fmt.Fprintf(output, "\n# reloaded files (took %v)", d)
}
}
}
func generateInstanceID() (string, error) {
return uuid.New(rand.Reader)
}
func generateDecisionID() string {
id, err := uuid.New(rand.Reader)
if err != nil {
return ""
}
return id
}
func init() {
registeredPlugins = make(map[string]plugins.Factory)
}