depot/packages/networking/ipfs-cluster/cmdutils/cmdutils.go

208 lines
4.9 KiB
Go
Raw Normal View History

2022-10-19 23:23:11 +03:00
// Package cmdutils contains utilities to facilitate building of command line
// applications launching cluster peers.
package cmdutils
import (
"context"
"fmt"
"io"
"net"
"os"
"os/signal"
"strings"
"syscall"
"time"
"github.com/ipfs/go-datastore"
ipfscluster "github.com/ipfs-cluster/ipfs-cluster"
ipfshttp "github.com/ipfs-cluster/ipfs-cluster/ipfsconn/ipfshttp"
host "github.com/libp2p/go-libp2p/core/host"
2022-10-19 23:23:11 +03:00
dual "github.com/libp2p/go-libp2p-kad-dht/dual"
ma "github.com/multiformats/go-multiaddr"
"github.com/pkg/errors"
"go.uber.org/multierr"
)
// RandomizePorts replaces TCP and UDP ports with random, but valid port
// values, on the given multiaddresses
func RandomizePorts(addrs []ma.Multiaddr) ([]ma.Multiaddr, error) {
results := make([]ma.Multiaddr, 0, len(addrs))
for _, m := range addrs {
var prev string
var err error
components := []ma.Multiaddr{}
ma.ForEach(m, func(c ma.Component) bool {
code := c.Protocol().Code
if code != ma.P_TCP && code != ma.P_UDP {
components = append(components, &c)
prev = c.Value()
return true
}
var ln io.Closer
var port int
ip := prev
if strings.Contains(ip, ":") { // ipv6 needs bracketing
ip = "[" + ip + "]"
}
if c.Protocol().Code == ma.P_UDP {
ln, port, err = listenUDP(c.Protocol().Name, ip)
} else {
ln, port, err = listenTCP(c.Protocol().Name, ip)
}
if err != nil {
return false
}
defer ln.Close()
var c1 *ma.Component
c1, err = ma.NewComponent(c.Protocol().Name, fmt.Sprintf("%d", port))
if err != nil {
return false
}
components = append(components, c1)
prev = c.Value()
return true
})
if err != nil {
return results, err
}
results = append(results, ma.Join(components...))
}
return results, nil
}
// returns the listener so it can be closed later and port
func listenTCP(name, ip string) (io.Closer, int, error) {
ln, err := net.Listen(name, ip+":0")
if err != nil {
return nil, 0, err
}
return ln, ln.Addr().(*net.TCPAddr).Port, nil
}
// returns the listener so it can be cloesd later and port
func listenUDP(name, ip string) (io.Closer, int, error) {
ln, err := net.ListenPacket(name, ip+":0")
if err != nil {
return nil, 0, err
}
return ln, ln.LocalAddr().(*net.UDPAddr).Port, nil
}
// HandleSignals orderly shuts down an IPFS Cluster peer
// on SIGINT, SIGTERM, SIGHUP. It forces command termination
// on the 3rd-signal count.
func HandleSignals(
ctx context.Context,
cancel context.CancelFunc,
cluster *ipfscluster.Cluster,
host host.Host,
dht *dual.DHT,
store datastore.Datastore,
) error {
signalChan := make(chan os.Signal, 20)
signal.Notify(
signalChan,
syscall.SIGINT,
syscall.SIGTERM,
syscall.SIGHUP,
)
var ctrlcCount int
for {
select {
case <-signalChan:
ctrlcCount++
handleCtrlC(ctx, cluster, ctrlcCount)
case <-cluster.Done():
cancel()
return multierr.Combine(
dht.Close(),
host.Close(),
store.Close(),
)
}
}
}
func handleCtrlC(ctx context.Context, cluster *ipfscluster.Cluster, ctrlcCount int) {
switch ctrlcCount {
case 1:
go func() {
if err := cluster.Shutdown(ctx); err != nil {
ErrorOut("error shutting down cluster: %s", err)
os.Exit(1)
}
}()
case 2:
ErrorOut(`
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Shutdown is taking too long! Press Ctrl-c again to manually kill cluster.
Note that this may corrupt the local cluster state.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
`)
case 3:
ErrorOut("exiting cluster NOW")
os.Exit(1)
}
}
// ErrorOut formats something and prints it to sdterr.
func ErrorOut(m string, a ...interface{}) {
fmt.Fprintf(os.Stderr, m, a...)
}
// WaitForIPFS hangs until IPFS API becomes available or the given context is
// canceled. The IPFS API location is determined by the default ipfshttp
// component configuration and can be overridden using environment variables
// that affect that configuration. Note that we have to do this in the blind,
// since we want to wait for IPFS before we even fetch the IPFS component
// configuration (because the configuration might be hosted on IPFS itself)
func WaitForIPFS(ctx context.Context) error {
ipfshttpCfg := ipfshttp.Config{}
ipfshttpCfg.Default()
ipfshttpCfg.ApplyEnvVars()
ipfshttpCfg.ConnectSwarmsDelay = 0
ipfshttpCfg.Tracing = false
ipfscluster.SetFacilityLogLevel("ipfshttp", "critical")
defer ipfscluster.SetFacilityLogLevel("ipfshttp", "info")
ipfs, err := ipfshttp.NewConnector(&ipfshttpCfg)
if err != nil {
return errors.Wrap(err, "error creating an ipfshttp instance to wait for IPFS")
}
i := 0
for {
select {
case <-ctx.Done():
return ctx.Err()
default:
if i%10 == 0 {
fmt.Printf("waiting for IPFS to become available on %s...\n", ipfshttpCfg.NodeAddr)
}
i++
time.Sleep(time.Second)
_, err := ipfs.ID(ctx)
if err == nil {
// sleep an extra second and quit
time.Sleep(time.Second)
return nil
}
}
}
}