Implement oboard diagnostics admin API (#9024)

- Implement a graph algorithm to test network bandwidth from every 
  node to every other node
- Saturate any network bandwidth adaptively, accounting for slow 
  and fast network capacity
- Implement parallel drive OBD tests
- Implement a paging mechanism for OBD test to provide periodic updates to client
- Implement Sys, Process, Host, Mem OBD Infos
This commit is contained in:
Sidhartha Mani
2020-03-26 21:07:39 -07:00
committed by GitHub
parent 2777956581
commit 0c80bf45d0
21 changed files with 2153 additions and 7 deletions
+286
View File
@@ -25,6 +25,7 @@ import (
"net"
"net/url"
"path"
"sort"
"strings"
"sync"
"time"
@@ -38,6 +39,7 @@ import (
objectlock "github.com/minio/minio/pkg/bucket/object/lock"
"github.com/minio/minio/pkg/bucket/policy"
"github.com/minio/minio-go/v6/pkg/set"
"github.com/minio/minio/pkg/event"
"github.com/minio/minio/pkg/madmin"
xnet "github.com/minio/minio/pkg/net"
@@ -890,6 +892,290 @@ func (sys *NotificationSys) CollectNetPerfInfo(size int64) map[string][]ServerNe
return reply
}
// NetOBDInfo - Net OBD information
func (sys *NotificationSys) NetOBDInfo(ctx context.Context) madmin.ServerNetOBDInfo {
var sortedGlobalEndpoints []string
/*
Ensure that only untraversed links are visited by this server
i.e. if netOBD tests have been performed between a -> b, then do
not run it between b -> a
The graph of tests looks like this
a b c d
a | o | x | x | x |
b | o | o | x | x |
c | o | o | o | x |
d | o | o | o | o |
'x's should be tested, and 'o's should be skipped
*/
stripPath := func(hostPath string) string {
return strings.Split(hostPath, slashSeparator)[0]
}
hostSet := set.NewStringSet()
for _, ez := range globalEndpoints {
for _, e := range ez.Endpoints {
host := stripPath(e.Host)
if hostSet.Contains(host) {
sortedGlobalEndpoints = append(sortedGlobalEndpoints, host)
hostSet.Add(host)
}
}
}
sort.Strings(sortedGlobalEndpoints)
var remoteTargets []*peerRESTClient
search := func(host string) *peerRESTClient {
for index, client := range sys.peerClients {
if client == nil {
continue
}
if sys.peerClients[index].host.String() == host {
return client
}
}
return nil
}
for i := 0; i < len(sortedGlobalEndpoints); i++ {
if sortedGlobalEndpoints[i] != GetLocalPeer(globalEndpoints) {
continue
}
for j := 0; j < len(sortedGlobalEndpoints); j++ {
if j > i {
remoteTarget := search(sortedGlobalEndpoints[j])
if remoteTarget != nil {
remoteTargets = append(remoteTargets, remoteTarget)
}
}
}
}
netOBDs := make([]madmin.NetOBDInfo, len(remoteTargets))
for index, client := range remoteTargets {
if client == nil {
continue
}
var err error
netOBDs[index], err = client.NetOBDInfo(ctx)
addr := client.host.String()
reqInfo := (&logger.ReqInfo{}).AppendTags("remotePeer", addr)
ctx := logger.SetReqInfo(context.Background(), reqInfo)
logger.LogIf(ctx, err)
netOBDs[index].Addr = addr
if err != nil {
netOBDs[index].Error = err.Error()
}
}
return madmin.ServerNetOBDInfo{
Net: netOBDs,
Addr: GetLocalPeer(globalEndpoints),
}
}
// DispatchNetOBDInfo - Net OBD information from other nodes
func (sys *NotificationSys) DispatchNetOBDInfo(ctx context.Context) []madmin.ServerNetOBDInfo {
serverNetOBDs := []madmin.ServerNetOBDInfo{}
for index, client := range sys.peerClients {
if client == nil {
continue
}
serverNetOBD, err := sys.peerClients[index].DispatchNetOBDInfo(ctx)
if err != nil {
serverNetOBD.Addr = client.host.String()
serverNetOBD.Error = err.Error()
}
serverNetOBDs = append(serverNetOBDs, serverNetOBD)
}
return serverNetOBDs
}
// DriveOBDInfo - Drive OBD information
func (sys *NotificationSys) DriveOBDInfo(ctx context.Context) []madmin.ServerDrivesOBDInfo {
reply := make([]madmin.ServerDrivesOBDInfo, len(sys.peerClients))
g := errgroup.WithNErrs(len(sys.peerClients))
for index, client := range sys.peerClients {
if client == nil {
continue
}
index := index
g.Go(func() error {
var err error
reply[index], err = sys.peerClients[index].DriveOBDInfo(ctx)
return err
}, index)
}
for index, err := range g.Wait() {
if err != nil {
addr := sys.peerClients[index].host.String()
reqInfo := (&logger.ReqInfo{}).AppendTags("remotePeer", addr)
ctx := logger.SetReqInfo(context.Background(), reqInfo)
logger.LogIf(ctx, err)
reply[index].Addr = addr
reply[index].Error = err.Error()
}
}
return reply
}
// CPUOBDInfo - CPU OBD information
func (sys *NotificationSys) CPUOBDInfo(ctx context.Context) []madmin.ServerCPUOBDInfo {
reply := make([]madmin.ServerCPUOBDInfo, len(sys.peerClients))
g := errgroup.WithNErrs(len(sys.peerClients))
for index, client := range sys.peerClients {
if client == nil {
continue
}
index := index
g.Go(func() error {
var err error
reply[index], err = sys.peerClients[index].CPUOBDInfo(ctx)
return err
}, index)
}
for index, err := range g.Wait() {
if err != nil {
addr := sys.peerClients[index].host.String()
reqInfo := (&logger.ReqInfo{}).AppendTags("remotePeer", addr)
ctx := logger.SetReqInfo(context.Background(), reqInfo)
logger.LogIf(ctx, err)
reply[index].Addr = addr
reply[index].Error = err.Error()
}
}
return reply
}
// DiskHwOBDInfo - Disk HW OBD information
func (sys *NotificationSys) DiskHwOBDInfo(ctx context.Context) []madmin.ServerDiskHwOBDInfo {
reply := make([]madmin.ServerDiskHwOBDInfo, len(sys.peerClients))
g := errgroup.WithNErrs(len(sys.peerClients))
for index, client := range sys.peerClients {
if client == nil {
continue
}
index := index
g.Go(func() error {
var err error
reply[index], err = sys.peerClients[index].DiskHwOBDInfo(ctx)
return err
}, index)
}
for index, err := range g.Wait() {
if err != nil {
addr := sys.peerClients[index].host.String()
reqInfo := (&logger.ReqInfo{}).AppendTags("remotePeer", addr)
ctx := logger.SetReqInfo(context.Background(), reqInfo)
logger.LogIf(ctx, err)
reply[index].Addr = addr
reply[index].Error = err.Error()
}
}
return reply
}
// OsOBDInfo - Os OBD information
func (sys *NotificationSys) OsOBDInfo(ctx context.Context) []madmin.ServerOsOBDInfo {
reply := make([]madmin.ServerOsOBDInfo, len(sys.peerClients))
g := errgroup.WithNErrs(len(sys.peerClients))
for index, client := range sys.peerClients {
if client == nil {
continue
}
index := index
g.Go(func() error {
var err error
reply[index], err = sys.peerClients[index].OsOBDInfo(ctx)
return err
}, index)
}
for index, err := range g.Wait() {
if err != nil {
addr := sys.peerClients[index].host.String()
reqInfo := (&logger.ReqInfo{}).AppendTags("remotePeer", addr)
ctx := logger.SetReqInfo(context.Background(), reqInfo)
logger.LogIf(ctx, err)
reply[index].Addr = addr
reply[index].Error = err.Error()
}
}
return reply
}
// MemOBDInfo - Mem OBD information
func (sys *NotificationSys) MemOBDInfo(ctx context.Context) []madmin.ServerMemOBDInfo {
reply := make([]madmin.ServerMemOBDInfo, len(sys.peerClients))
g := errgroup.WithNErrs(len(sys.peerClients))
for index, client := range sys.peerClients {
if client == nil {
continue
}
index := index
g.Go(func() error {
var err error
reply[index], err = sys.peerClients[index].MemOBDInfo(ctx)
return err
}, index)
}
for index, err := range g.Wait() {
if err != nil {
addr := sys.peerClients[index].host.String()
reqInfo := (&logger.ReqInfo{}).AppendTags("remotePeer", addr)
ctx := logger.SetReqInfo(context.Background(), reqInfo)
logger.LogIf(ctx, err)
reply[index].Addr = addr
reply[index].Error = err.Error()
}
}
return reply
}
// ProcOBDInfo - Process OBD information
func (sys *NotificationSys) ProcOBDInfo(ctx context.Context) []madmin.ServerProcOBDInfo {
reply := make([]madmin.ServerProcOBDInfo, len(sys.peerClients))
g := errgroup.WithNErrs(len(sys.peerClients))
for index, client := range sys.peerClients {
if client == nil {
continue
}
index := index
g.Go(func() error {
var err error
reply[index], err = sys.peerClients[index].ProcOBDInfo(ctx)
return err
}, index)
}
for index, err := range g.Wait() {
if err != nil {
addr := sys.peerClients[index].host.String()
reqInfo := (&logger.ReqInfo{}).AppendTags("remotePeer", addr)
ctx := logger.SetReqInfo(context.Background(), reqInfo)
logger.LogIf(ctx, err)
reply[index].Addr = addr
reply[index].Error = err.Error()
}
}
return reply
}
// DrivePerfInfo - Drive speed (read and write) information
func (sys *NotificationSys) DrivePerfInfo(size int64) []madmin.ServerDrivesPerfInfo {
reply := make([]madmin.ServerDrivesPerfInfo, len(sys.peerClients))