tcp: use GetExtendedTcpTable to display states Closed, Listening, SynSent, SynRcvd, CloseWait, TimeWait ... (#1638)

Co-authored-by: Jan-Otto Kröpke <github@jkroepke.de>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Jan-Otto Kröpke <joe@cloudeteer.de>
This commit is contained in:
astigmata 2024-10-01 23:23:35 +02:00 committed by GitHub
parent 14910efd4f
commit 1caef5cc12
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
6 changed files with 215 additions and 8 deletions

View File

@ -26,6 +26,7 @@ Name | Description | Type | Labels
`windows_tcp_segments_received_total` | Total segments received, including those received in error. This count includes segments received on currently established connections | counter | af
`windows_tcp_segments_retransmitted_total` | Total segments retransmitted. That is, segments transmitted that contain one or more previously transmitted bytes | counter | af
`windows_tcp_segments_sent_total` | Total segments sent, including those on current connections, but excluding those containing *only* retransmitted bytes | counter | af
`windows_tcp_connections_state_count` | Number of TCP connections by state among: CLOSED, LISTENING, SYN_SENT, SYN_RECEIVED, ESTABLISHED, FIN_WAIT1, FIN_WAIT2, CLOSE_WAIT, CLOSING, LAST_ACK, TIME_WAIT, DELETE_TCB | gauge | af
### Example metric
_This collector does not yet have explained examples, we would appreciate your help adding them!_

View File

@ -3,6 +3,7 @@ package tcp
// Win32_PerfRawData_Tcpip_TCPv4 docs
// - https://msdn.microsoft.com/en-us/library/aa394341(v=vs.85).aspx
// The TCPv6 performance object uses the same fields.
// https://learn.microsoft.com/en-us/dotnet/api/system.net.networkinformation.tcpstate?view=net-8.0.
const (
connectionFailures = "Connection Failures"
connectionsActive = "Connections Active"

View File

@ -7,10 +7,12 @@ import (
"log/slog"
"github.com/alecthomas/kingpin/v2"
"github.com/prometheus-community/windows_exporter/pkg/headers/iphlpapi"
"github.com/prometheus-community/windows_exporter/pkg/perfdata"
"github.com/prometheus-community/windows_exporter/pkg/types"
"github.com/prometheus/client_golang/prometheus"
"github.com/yusufpapurcu/wmi"
"golang.org/x/sys/windows"
)
const Name = "tcp"
@ -35,6 +37,7 @@ type Collector struct {
segmentsReceivedTotal *prometheus.Desc
segmentsRetransmittedTotal *prometheus.Desc
segmentsSentTotal *prometheus.Desc
connectionsStateCount *prometheus.Desc
}
func New(config *Config) *Collector {
@ -144,6 +147,11 @@ func (c *Collector) Build(_ *slog.Logger, _ *wmi.Client) error {
[]string{"af"},
nil,
)
c.connectionsStateCount = prometheus.NewDesc(
prometheus.BuildFQName(types.Namespace, Name, "connections_state_count"),
"Number of TCP connections by state and address family",
[]string{"af", "state"}, nil,
)
return nil
}
@ -160,10 +168,36 @@ func (c *Collector) Collect(_ *types.ScrapeContext, logger *slog.Logger, ch chan
return err
}
if err := c.collectConnectionsState(ch); err != nil {
logger.Error("failed collecting tcp connection state metrics",
slog.Any("err", err),
)
return err
}
return nil
}
func writeTCPCounters(metrics map[string]perfdata.CounterValues, labels []string, c *Collector, ch chan<- prometheus.Metric) {
func (c *Collector) collect(ch chan<- prometheus.Metric) error {
data, err := c.perfDataCollector4.Collect()
if err != nil {
return fmt.Errorf("failed to collect TCPv4 metrics: %w", err)
}
c.writeTCPCounters(ch, data[perfdata.EmptyInstance], []string{"ipv4"})
data, err = c.perfDataCollector6.Collect()
if err != nil {
return fmt.Errorf("failed to collect TCPv6 metrics: %w", err)
}
c.writeTCPCounters(ch, data[perfdata.EmptyInstance], []string{"ipv6"})
return nil
}
func (c *Collector) writeTCPCounters(ch chan<- prometheus.Metric, metrics map[string]perfdata.CounterValues, labels []string) {
ch <- prometheus.MustNewConstMetric(
c.connectionFailures,
prometheus.CounterValue,
@ -220,20 +254,32 @@ func writeTCPCounters(metrics map[string]perfdata.CounterValues, labels []string
)
}
func (c *Collector) collect(ch chan<- prometheus.Metric) error {
data, err := c.perfDataCollector4.Collect()
func (c *Collector) collectConnectionsState(ch chan<- prometheus.Metric) error {
stateCounts, err := iphlpapi.GetTCPConnectionStates(windows.AF_INET)
if err != nil {
return fmt.Errorf("failed to collect TCPv4 metrics: %w", err)
return fmt.Errorf("failed to collect TCP connection states for %s: %w", "ipv4", err)
}
writeTCPCounters(data[perfdata.EmptyInstance], []string{"ipv4"}, c, ch)
c.sendTCPStateMetrics(ch, stateCounts, "ipv4")
data, err = c.perfDataCollector6.Collect()
stateCounts, err = iphlpapi.GetTCPConnectionStates(windows.AF_INET6)
if err != nil {
return fmt.Errorf("failed to collect TCPv6 metrics: %w", err)
return fmt.Errorf("failed to collect TCP6 connection states for %s: %w", "ipv6", err)
}
writeTCPCounters(data[perfdata.EmptyInstance], []string{"ipv6"}, c, ch)
c.sendTCPStateMetrics(ch, stateCounts, "ipv6")
return nil
}
func (c *Collector) sendTCPStateMetrics(ch chan<- prometheus.Metric, stateCounts map[iphlpapi.MIB_TCP_STATE]uint32, af string) {
for state, count := range stateCounts {
ch <- prometheus.MustNewConstMetric(
c.connectionsStateCount,
prometheus.GaugeValue,
float64(count),
af,
state.String(),
)
}
}

View File

@ -0,0 +1,6 @@
package iphlpapi
const (
TCPTableClass uint32 = 5
TCP6TableClass uint32 = 5
)

View File

@ -0,0 +1,77 @@
package iphlpapi
import (
"fmt"
"unsafe"
"golang.org/x/sys/windows"
)
var (
modiphlpapi = windows.NewLazySystemDLL("iphlpapi.dll")
procGetExtendedTcpTable = modiphlpapi.NewProc("GetExtendedTcpTable")
)
func GetTCPConnectionStates(family uint32) (map[MIB_TCP_STATE]uint32, error) {
var size uint32
stateCounts := make(map[MIB_TCP_STATE]uint32)
rowSize := uint32(unsafe.Sizeof(MIB_TCPROW_OWNER_PID{}))
tableClass := TCPTableClass
if family == windows.AF_INET6 {
rowSize = uint32(unsafe.Sizeof(MIB_TCP6ROW_OWNER_PID{}))
tableClass = TCP6TableClass
}
ret := getExtendedTcpTable(0, &size, true, family, tableClass, 0)
if ret != 0 && ret != uintptr(windows.ERROR_INSUFFICIENT_BUFFER) {
return nil, fmt.Errorf("getExtendedTcpTable (size query) failed with code %d", ret)
}
buf := make([]byte, size)
ret = getExtendedTcpTable(uintptr(unsafe.Pointer(&buf[0])), &size, true, family, tableClass, 0)
if ret != 0 {
return nil, fmt.Errorf("getExtendedTcpTable (data query) failed with code %d", ret)
}
numEntries := *(*uint32)(unsafe.Pointer(&buf[0]))
for i := range numEntries {
var state MIB_TCP_STATE
if family == windows.AF_INET6 {
row := (*MIB_TCP6ROW_OWNER_PID)(unsafe.Pointer(&buf[4+i*rowSize]))
state = row.dwState
} else {
row := (*MIB_TCPROW_OWNER_PID)(unsafe.Pointer(&buf[4+i*rowSize]))
state = row.dwState
}
stateCounts[state]++
}
return stateCounts, nil
}
func getExtendedTcpTable(pTCPTable uintptr, pdwSize *uint32, bOrder bool, ulAf uint32, tableClass uint32, reserved uint32) uintptr {
ret, _, _ := procGetExtendedTcpTable.Call(
pTCPTable,
uintptr(unsafe.Pointer(pdwSize)),
uintptr(boolToInt(bOrder)),
uintptr(ulAf),
uintptr(tableClass),
uintptr(reserved),
)
return ret
}
func boolToInt(b bool) int {
if b {
return 1
}
return 0
}

View File

@ -0,0 +1,76 @@
package iphlpapi
import "fmt"
// MIB_TCPROW_OWNER_PID structure for IPv4.
// https://learn.microsoft.com/en-us/windows/win32/api/tcpmib/ns-tcpmib-mib_tcprow_owner_pid
type MIB_TCPROW_OWNER_PID struct {
dwState MIB_TCP_STATE
dwLocalAddr uint32
dwLocalPort uint32
dwRemoteAddr uint32
dwRemotePort uint32
dwOwningPid uint32
}
// MIB_TCP6ROW_OWNER_PID structure for IPv6.
// https://learn.microsoft.com/en-us/windows/win32/api/tcpmib/ns-tcpmib-mib_tcp6row_owner_pid
type MIB_TCP6ROW_OWNER_PID struct {
ucLocalAddr [16]byte
dwLocalScopeId uint32
dwLocalPort uint32
ucRemoteAddr [16]byte
dwRemoteScopeId uint32
dwRemotePort uint32
dwState MIB_TCP_STATE
dwOwningPid uint32
}
type MIB_TCP_STATE uint32
const (
_ MIB_TCP_STATE = iota
TCPStateClosed
TCPStateListening
TCPStateSynSent
TCPStateSynRcvd
TCPStateEstablished
TCPStateFinWait1
TCPStateFinWait2
TCPStateCloseWait
TCPStateClosing
TCPStateLastAck
TCPStateTimeWait
TCPStateDeleteTcb
)
func (state MIB_TCP_STATE) String() string {
switch state {
case TCPStateClosed:
return "CLOSED"
case TCPStateListening:
return "LISTENING"
case TCPStateSynSent:
return "SYN_SENT"
case TCPStateSynRcvd:
return "SYN_RECEIVED"
case TCPStateEstablished:
return "ESTABLISHED"
case TCPStateFinWait1:
return "FIN_WAIT1"
case TCPStateFinWait2:
return "FIN_WAIT2"
case TCPStateCloseWait:
return "CLOSE_WAIT"
case TCPStateClosing:
return "CLOSING"
case TCPStateLastAck:
return "LAST_ACK"
case TCPStateTimeWait:
return "TIME_WAIT"
case TCPStateDeleteTcb:
return "DELETE_TCB"
default:
return fmt.Sprintf("UNKNOWN_%d", state)
}
}