2018-01-09 23:26:53 +00:00
|
|
|
// Package servers provides an interface for choosing Servers to communicate
|
|
|
|
// with from a Nomad Client perspective. The package does not provide any API
|
|
|
|
// guarantees and should be called only by `hashicorp/nomad`.
|
|
|
|
package servers
|
|
|
|
|
|
|
|
import (
|
|
|
|
"math/rand"
|
|
|
|
"net"
|
2018-04-05 17:40:17 +00:00
|
|
|
"sort"
|
2018-01-09 23:26:53 +00:00
|
|
|
"strings"
|
|
|
|
"sync"
|
|
|
|
"time"
|
|
|
|
|
|
|
|
"github.com/hashicorp/consul/lib"
|
2018-08-29 22:05:03 +00:00
|
|
|
hclog "github.com/hashicorp/go-hclog"
|
2018-01-09 23:26:53 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
const (
|
|
|
|
// clientRPCMinReuseDuration controls the minimum amount of time RPC
|
|
|
|
// queries are sent over an established connection to a single server
|
2018-01-10 19:01:46 +00:00
|
|
|
clientRPCMinReuseDuration = 5 * time.Minute
|
2018-01-09 23:26:53 +00:00
|
|
|
|
|
|
|
// Limit the number of new connections a server receives per second
|
|
|
|
// for connection rebalancing. This limit caps the load caused by
|
|
|
|
// continual rebalancing efforts when a cluster is in equilibrium. A
|
|
|
|
// lower value comes at the cost of increased recovery time after a
|
|
|
|
// partition. This parameter begins to take effect when there are
|
|
|
|
// more than ~48K clients querying 5x servers or at lower server
|
|
|
|
// values when there is a partition.
|
|
|
|
//
|
|
|
|
// For example, in a 100K Nomad cluster with 5x servers, it will
|
|
|
|
// take ~5min for all servers to rebalance their connections. If
|
|
|
|
// 99,995 agents are in the minority talking to only one server, it
|
|
|
|
// will take ~26min for all servers to rebalance. A 10K cluster in
|
|
|
|
// the same scenario will take ~2.6min to rebalance.
|
|
|
|
newRebalanceConnsPerSecPerServer = 64
|
|
|
|
)
|
|
|
|
|
2018-01-25 05:14:03 +00:00
|
|
|
// Pinger is an interface for pinging a server to see if it is healthy.
|
|
|
|
type Pinger interface {
|
|
|
|
Ping(addr net.Addr) error
|
|
|
|
}
|
|
|
|
|
2018-01-09 23:26:53 +00:00
|
|
|
// Server contains the address of a server and metadata that can be used for
|
|
|
|
// choosing a server to contact.
|
|
|
|
type Server struct {
|
|
|
|
// Addr is the resolved address of the server
|
|
|
|
Addr net.Addr
|
|
|
|
addr string
|
2018-01-25 02:00:21 +00:00
|
|
|
sync.Mutex
|
2018-01-09 23:26:53 +00:00
|
|
|
|
|
|
|
// DC is the datacenter of the server
|
|
|
|
DC string
|
|
|
|
}
|
|
|
|
|
2018-01-25 17:48:52 +00:00
|
|
|
func (s *Server) Copy() *Server {
|
2018-01-26 01:56:47 +00:00
|
|
|
s.Lock()
|
|
|
|
defer s.Unlock()
|
|
|
|
|
2018-01-25 17:48:52 +00:00
|
|
|
return &Server{
|
|
|
|
Addr: s.Addr,
|
|
|
|
addr: s.addr,
|
|
|
|
DC: s.DC,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-01-09 23:26:53 +00:00
|
|
|
func (s *Server) String() string {
|
2018-01-25 02:00:21 +00:00
|
|
|
s.Lock()
|
|
|
|
defer s.Unlock()
|
|
|
|
|
2018-01-09 23:26:53 +00:00
|
|
|
if s.addr == "" {
|
|
|
|
s.addr = s.Addr.String()
|
|
|
|
}
|
|
|
|
|
|
|
|
return s.addr
|
|
|
|
}
|
|
|
|
|
2018-04-04 00:27:34 +00:00
|
|
|
func (s *Server) Equal(o *Server) bool {
|
|
|
|
if s == nil && o == nil {
|
|
|
|
return true
|
|
|
|
} else if s == nil && o != nil || s != nil && o == nil {
|
|
|
|
return false
|
|
|
|
}
|
|
|
|
|
|
|
|
return s.Addr.String() == o.Addr.String() && s.DC == o.DC
|
|
|
|
}
|
|
|
|
|
2018-01-09 23:26:53 +00:00
|
|
|
type Servers []*Server
|
|
|
|
|
|
|
|
func (s Servers) String() string {
|
|
|
|
addrs := make([]string, 0, len(s))
|
|
|
|
for _, srv := range s {
|
|
|
|
addrs = append(addrs, srv.String())
|
|
|
|
}
|
|
|
|
return strings.Join(addrs, ",")
|
|
|
|
}
|
|
|
|
|
2018-01-25 05:14:03 +00:00
|
|
|
// cycle cycles a list of servers in-place
|
|
|
|
func (s Servers) cycle() {
|
|
|
|
numServers := len(s)
|
2018-01-25 02:00:21 +00:00
|
|
|
if numServers < 2 {
|
2018-01-25 05:14:03 +00:00
|
|
|
return // No action required
|
2018-01-25 02:00:21 +00:00
|
|
|
}
|
|
|
|
|
2018-01-25 05:14:03 +00:00
|
|
|
start := s[0]
|
|
|
|
for i := 1; i < numServers; i++ {
|
|
|
|
s[i-1] = s[i]
|
|
|
|
}
|
|
|
|
s[numServers-1] = start
|
2018-01-25 02:00:21 +00:00
|
|
|
}
|
|
|
|
|
2018-01-25 05:14:03 +00:00
|
|
|
// shuffle shuffles the server list in place
|
|
|
|
func (s Servers) shuffle() {
|
|
|
|
for i := len(s) - 1; i > 0; i-- {
|
2018-01-25 02:00:21 +00:00
|
|
|
j := rand.Int31n(int32(i + 1))
|
2018-01-25 05:14:03 +00:00
|
|
|
s[i], s[j] = s[j], s[i]
|
2018-01-25 02:00:21 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-04-05 17:40:17 +00:00
|
|
|
func (s Servers) Sort() {
|
|
|
|
sort.Slice(s, func(i, j int) bool {
|
|
|
|
a, b := s[i], s[j]
|
|
|
|
if addr1, addr2 := a.Addr.String(), b.Addr.String(); addr1 == addr2 {
|
|
|
|
return a.DC < b.DC
|
|
|
|
} else {
|
|
|
|
return addr1 < addr2
|
|
|
|
}
|
|
|
|
})
|
|
|
|
}
|
|
|
|
|
|
|
|
// Equal returns if the two server lists are equal, including the ordering.
|
|
|
|
func (s Servers) Equal(o Servers) bool {
|
|
|
|
if len(s) != len(o) {
|
|
|
|
return false
|
|
|
|
}
|
|
|
|
|
|
|
|
for i, v := range s {
|
|
|
|
if !v.Equal(o[i]) {
|
|
|
|
return false
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return true
|
|
|
|
}
|
|
|
|
|
2018-01-09 23:26:53 +00:00
|
|
|
type Manager struct {
|
2018-01-25 05:14:03 +00:00
|
|
|
// servers is the list of all known Nomad servers.
|
|
|
|
servers Servers
|
2018-01-09 23:26:53 +00:00
|
|
|
|
|
|
|
// rebalanceTimer controls the duration of the rebalance interval
|
|
|
|
rebalanceTimer *time.Timer
|
|
|
|
|
|
|
|
// shutdownCh is a copy of the channel in Nomad.Client
|
|
|
|
shutdownCh chan struct{}
|
|
|
|
|
|
|
|
// numNodes is used to estimate the approximate number of nodes in
|
|
|
|
// a cluster and limit the rate at which it rebalances server
|
|
|
|
// connections. This should be read and set using atomic.
|
|
|
|
numNodes int32
|
|
|
|
|
|
|
|
// connPoolPinger is used to test the health of a server in the connection
|
|
|
|
// pool. Pinger is an interface that wraps client.ConnPool.
|
|
|
|
connPoolPinger Pinger
|
|
|
|
|
2018-08-29 22:05:03 +00:00
|
|
|
logger hclog.Logger
|
|
|
|
|
2018-01-25 05:14:03 +00:00
|
|
|
sync.Mutex
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// New is the only way to safely create a new Manager struct.
|
2018-08-29 22:05:03 +00:00
|
|
|
func New(logger hclog.Logger, shutdownCh chan struct{}, connPoolPinger Pinger) (m *Manager) {
|
|
|
|
logger = logger.Named("server_mgr")
|
2018-01-25 05:14:03 +00:00
|
|
|
return &Manager{
|
|
|
|
logger: logger,
|
|
|
|
connPoolPinger: connPoolPinger,
|
|
|
|
rebalanceTimer: time.NewTimer(clientRPCMinReuseDuration),
|
|
|
|
shutdownCh: shutdownCh,
|
|
|
|
}
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// Start is used to start and manage the task of automatically shuffling and
|
2018-01-25 05:14:03 +00:00
|
|
|
// rebalancing the list of Nomad servers in order to distribute load across
|
|
|
|
// all known and available Nomad servers.
|
2018-01-09 23:26:53 +00:00
|
|
|
func (m *Manager) Start() {
|
|
|
|
for {
|
|
|
|
select {
|
|
|
|
case <-m.rebalanceTimer.C:
|
|
|
|
m.RebalanceServers()
|
|
|
|
m.refreshServerRebalanceTimer()
|
|
|
|
|
|
|
|
case <-m.shutdownCh:
|
2018-08-29 22:05:03 +00:00
|
|
|
m.logger.Debug("shutting down")
|
2018-01-09 23:26:53 +00:00
|
|
|
return
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-04-05 17:40:17 +00:00
|
|
|
// SetServers sets the servers and returns if the new server list is different
|
|
|
|
// than the existing server set
|
|
|
|
func (m *Manager) SetServers(servers Servers) bool {
|
2018-01-25 05:14:03 +00:00
|
|
|
m.Lock()
|
|
|
|
defer m.Unlock()
|
2018-04-04 00:27:34 +00:00
|
|
|
|
2018-04-05 17:40:17 +00:00
|
|
|
// Sort both the existing and incoming servers
|
|
|
|
servers.Sort()
|
|
|
|
m.servers.Sort()
|
|
|
|
|
|
|
|
// Determine if they are equal
|
|
|
|
equal := servers.Equal(m.servers)
|
|
|
|
|
2018-04-04 00:27:34 +00:00
|
|
|
// Randomize the incoming servers
|
|
|
|
servers.shuffle()
|
2018-01-25 05:14:03 +00:00
|
|
|
m.servers = servers
|
2018-04-05 17:40:17 +00:00
|
|
|
|
|
|
|
return !equal
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
2018-01-25 05:14:03 +00:00
|
|
|
// FindServer returns a server to send an RPC too. If there are no servers, nil
|
|
|
|
// is returned.
|
2018-01-09 23:26:53 +00:00
|
|
|
func (m *Manager) FindServer() *Server {
|
2018-01-25 05:14:03 +00:00
|
|
|
m.Lock()
|
|
|
|
defer m.Unlock()
|
|
|
|
|
|
|
|
if len(m.servers) == 0 {
|
2018-08-29 22:05:03 +00:00
|
|
|
m.logger.Warn("no servers available")
|
2018-01-09 23:26:53 +00:00
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// Return whatever is at the front of the list because it is
|
|
|
|
// assumed to be the oldest in the server list (unless -
|
|
|
|
// hypothetically - the server list was rotated right after a
|
|
|
|
// server was added).
|
2018-01-25 05:14:03 +00:00
|
|
|
return m.servers[0]
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// NumNodes returns the number of approximate nodes in the cluster.
|
|
|
|
func (m *Manager) NumNodes() int32 {
|
2018-01-25 05:14:03 +00:00
|
|
|
m.Lock()
|
|
|
|
defer m.Unlock()
|
|
|
|
return m.numNodes
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// SetNumNodes stores the number of approximate nodes in the cluster.
|
|
|
|
func (m *Manager) SetNumNodes(n int32) {
|
2018-01-25 05:14:03 +00:00
|
|
|
m.Lock()
|
|
|
|
defer m.Unlock()
|
|
|
|
m.numNodes = n
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// NotifyFailedServer marks the passed in server as "failed" by rotating it
|
|
|
|
// to the end of the server list.
|
|
|
|
func (m *Manager) NotifyFailedServer(s *Server) {
|
2018-01-25 05:14:03 +00:00
|
|
|
m.Lock()
|
|
|
|
defer m.Unlock()
|
2018-01-09 23:26:53 +00:00
|
|
|
|
|
|
|
// If the server being failed is not the first server on the list,
|
|
|
|
// this is a noop. If, however, the server is failed and first on
|
2018-01-25 05:14:03 +00:00
|
|
|
// the list, move the server to the end of the list.
|
2018-04-04 00:27:34 +00:00
|
|
|
if len(m.servers) > 1 && m.servers[0].Equal(s) {
|
2018-01-25 05:14:03 +00:00
|
|
|
m.servers.cycle()
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-01-25 02:00:21 +00:00
|
|
|
// NumServers returns the total number of known servers whether healthy or not.
|
2018-01-09 23:26:53 +00:00
|
|
|
func (m *Manager) NumServers() int {
|
2018-01-25 05:14:03 +00:00
|
|
|
m.Lock()
|
|
|
|
defer m.Unlock()
|
|
|
|
return len(m.servers)
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// GetServers returns a copy of the current list of servers.
|
2018-01-25 05:14:03 +00:00
|
|
|
func (m *Manager) GetServers() Servers {
|
|
|
|
m.Lock()
|
|
|
|
defer m.Unlock()
|
|
|
|
|
|
|
|
copy := make([]*Server, 0, len(m.servers))
|
|
|
|
for _, s := range m.servers {
|
2018-01-25 17:48:52 +00:00
|
|
|
copy = append(copy, s.Copy())
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return copy
|
|
|
|
}
|
|
|
|
|
2018-01-25 05:14:03 +00:00
|
|
|
// RebalanceServers shuffles the order in which Servers will be contacted. The
|
|
|
|
// function will shuffle the set of potential servers to contact and then attempt
|
|
|
|
// to contact each server. If a server successfully responds it is used, otherwise
|
|
|
|
// it is rotated such that it will be the last attempted server.
|
2018-01-09 23:26:53 +00:00
|
|
|
func (m *Manager) RebalanceServers() {
|
|
|
|
// Shuffle servers so we have a chance of picking a new one.
|
2018-01-26 01:56:47 +00:00
|
|
|
servers := m.GetServers()
|
|
|
|
servers.shuffle()
|
2018-01-09 23:26:53 +00:00
|
|
|
|
|
|
|
// Iterate through the shuffled server list to find an assumed
|
|
|
|
// healthy server. NOTE: Do not iterate on the list directly because
|
|
|
|
// this loop mutates the server list in-place.
|
|
|
|
var foundHealthyServer bool
|
2018-01-25 05:14:03 +00:00
|
|
|
for i := 0; i < len(m.servers); i++ {
|
2018-01-09 23:26:53 +00:00
|
|
|
// Always test the first server. Failed servers are cycled
|
|
|
|
// while Serf detects the node has failed.
|
2018-01-26 01:56:47 +00:00
|
|
|
srv := servers[0]
|
2018-01-09 23:26:53 +00:00
|
|
|
|
2018-01-25 02:00:21 +00:00
|
|
|
err := m.connPoolPinger.Ping(srv.Addr)
|
|
|
|
if err == nil {
|
2018-01-09 23:26:53 +00:00
|
|
|
foundHealthyServer = true
|
|
|
|
break
|
|
|
|
}
|
2018-08-29 22:05:03 +00:00
|
|
|
m.logger.Debug("error pinging server", "error", err, "server", srv)
|
2018-01-09 23:26:53 +00:00
|
|
|
|
2018-01-26 01:56:47 +00:00
|
|
|
servers.cycle()
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
2018-01-25 02:00:21 +00:00
|
|
|
if !foundHealthyServer {
|
2018-08-29 22:05:03 +00:00
|
|
|
m.logger.Debug("no healthy servers during rebalance")
|
2018-01-26 01:56:47 +00:00
|
|
|
return
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
2018-01-26 01:56:47 +00:00
|
|
|
// Save the servers
|
|
|
|
m.Lock()
|
|
|
|
m.servers = servers
|
|
|
|
m.Unlock()
|
2018-01-09 23:26:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// refreshServerRebalanceTimer is only called once m.rebalanceTimer expires.
|
|
|
|
func (m *Manager) refreshServerRebalanceTimer() time.Duration {
|
2018-01-25 05:14:03 +00:00
|
|
|
m.Lock()
|
|
|
|
defer m.Unlock()
|
|
|
|
numServers := len(m.servers)
|
2018-01-10 19:01:46 +00:00
|
|
|
|
2018-01-09 23:26:53 +00:00
|
|
|
// Limit this connection's life based on the size (and health) of the
|
|
|
|
// cluster. Never rebalance a connection more frequently than
|
|
|
|
// connReuseLowWatermarkDuration, and make sure we never exceed
|
|
|
|
// clusterWideRebalanceConnsPerSec operations/s across numLANMembers.
|
|
|
|
clusterWideRebalanceConnsPerSec := float64(numServers * newRebalanceConnsPerSecPerServer)
|
|
|
|
|
2018-01-25 05:14:03 +00:00
|
|
|
connRebalanceTimeout := lib.RateScaledInterval(clusterWideRebalanceConnsPerSec, clientRPCMinReuseDuration, int(m.numNodes))
|
2018-01-09 23:26:53 +00:00
|
|
|
connRebalanceTimeout += lib.RandomStagger(connRebalanceTimeout)
|
|
|
|
|
|
|
|
m.rebalanceTimer.Reset(connRebalanceTimeout)
|
|
|
|
return connRebalanceTimeout
|
|
|
|
}
|
|
|
|
|
|
|
|
// ResetRebalanceTimer resets the rebalance timer. This method exists for
|
|
|
|
// testing and should not be used directly.
|
|
|
|
func (m *Manager) ResetRebalanceTimer() {
|
2018-01-25 05:14:03 +00:00
|
|
|
m.Lock()
|
|
|
|
defer m.Unlock()
|
2018-01-09 23:26:53 +00:00
|
|
|
m.rebalanceTimer.Reset(clientRPCMinReuseDuration)
|
|
|
|
}
|