Removed heal backend (#3188)

Signed-off-by: Benjamin Perez <benjamin@bexsoft.net>
This commit is contained in:
Alex
2024-01-10 23:16:08 -06:00
committed by GitHub
parent 08c922dca6
commit 343ff575e6
5 changed files with 0 additions and 692 deletions

View File

@@ -1,375 +0,0 @@
// This file is part of MinIO Console Server
// Copyright (c) 2021 MinIO, Inc.
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package api
import (
"context"
"encoding/json"
"fmt"
"net/http"
"regexp"
"strconv"
"strings"
"time"
"github.com/minio/madmin-go/v3"
"github.com/minio/websocket"
)
// An alias of string to represent the health color code of an object
type col string
const (
colGrey col = "Grey"
colRed col = "Red"
colYellow col = "Yellow"
colGreen col = "Green"
)
var (
hColOrder = []col{colRed, colYellow, colGreen}
hColTable = map[int][]int{
1: {0, -1, 1},
2: {0, 1, 2},
3: {1, 2, 3},
4: {1, 2, 4},
5: {1, 3, 5},
6: {2, 4, 6},
7: {2, 4, 7},
8: {2, 5, 8},
}
)
type healItemStatus struct {
Status string `json:"status"`
Error string `json:"errors,omitempty"`
Type string `json:"type"`
Name string `json:"name"`
Before struct {
Color string `json:"color"`
Offline int `json:"offline"`
Online int `json:"online"`
Missing int `json:"missing"`
Corrupted int `json:"corrupted"`
Drives []madmin.HealDriveInfo `json:"drives"`
} `json:"before"`
After struct {
Color string `json:"color"`
Offline int `json:"offline"`
Online int `json:"online"`
Missing int `json:"missing"`
Corrupted int `json:"corrupted"`
Drives []madmin.HealDriveInfo `json:"drives"`
} `json:"after"`
Size int64 `json:"size"`
}
type healStatus struct {
// Total time since heal start in seconds
HealDuration float64 `json:"healDuration"`
// Accumulated statistics of heal result records
BytesScanned int64 `json:"bytesScanned"`
// Counter for objects, and another counter for all kinds of
// items
ObjectsScanned int64 `json:"objectsScanned"`
ItemsScanned int64 `json:"itemsScanned"`
// Counters for healed objects and all kinds of healed items
ObjectsHealed int64 `json:"objectsHealed"`
ItemsHealed int64 `json:"itemsHealed"`
ItemsHealthStatus []healItemStatus `json:"itemsHealthStatus"`
// Map of health color code to number of objects with that
// health color code.
HealthBeforeCols map[col]int64 `json:"healthBeforeCols"`
HealthAfterCols map[col]int64 `json:"healthAfterCols"`
}
type healOptions struct {
BucketName string
Prefix string
ForceStart bool
ForceStop bool
madmin.HealOpts
}
// startHeal starts healing of the servers based on heal options
func startHeal(ctx context.Context, conn WSConn, client MinioAdmin, hOpts *healOptions) error {
// Initialize heal
healStart, _, err := client.heal(ctx, hOpts.BucketName, hOpts.Prefix, hOpts.HealOpts, "", hOpts.ForceStart, hOpts.ForceStop)
if err != nil {
LogError("error initializing healing: %v", err)
return err
}
if hOpts.ForceStop {
return nil
}
clientToken := healStart.ClientToken
hs := healStatus{
HealthBeforeCols: make(map[col]int64),
HealthAfterCols: make(map[col]int64),
}
for {
select {
case <-ctx.Done():
return nil
default:
_, res, err := client.heal(ctx, hOpts.BucketName, hOpts.Prefix, hOpts.HealOpts, clientToken, hOpts.ForceStart, hOpts.ForceStop)
if err != nil {
LogError("error on heal: %v", err)
return err
}
hs.writeStatus(&res, conn)
if res.Summary == "finished" {
return nil
}
if res.Summary == "stopped" {
return fmt.Errorf("heal had an errors - %s", res.FailureDetail)
}
time.Sleep(time.Second)
}
}
}
func (h *healStatus) writeStatus(s *madmin.HealTaskStatus, conn WSConn) error {
// Update state
h.updateDuration(s)
for _, item := range s.Items {
err := h.updateStats(item)
if err != nil {
LogError("error on updateStats: %v", err)
return err
}
}
// Serialize message to be sent
infoBytes, err := json.Marshal(h)
if err != nil {
LogError("error on json.Marshal: %v", err)
return err
}
// Send Message through websocket connection
err = conn.writeMessage(websocket.TextMessage, infoBytes)
if err != nil {
LogError("error writeMessage: %v", err)
return err
}
return nil
}
func (h *healStatus) updateDuration(s *madmin.HealTaskStatus) {
h.HealDuration = time.Now().UTC().Sub(s.StartTime).Round(time.Second).Seconds()
}
func (h *healStatus) updateStats(i madmin.HealResultItem) error {
// update general status
if i.Type == madmin.HealItemObject {
// Objects whose size could not be found have -1 size
// returned.
if i.ObjectSize >= 0 {
h.BytesScanned += i.ObjectSize
}
h.ObjectsScanned++
}
h.ItemsScanned++
beforeUp, afterUp := i.GetOnlineCounts()
if afterUp > beforeUp {
if i.Type == madmin.HealItemObject {
h.ObjectsHealed++
}
h.ItemsHealed++
}
// update per item status
itemStatus := healItemStatus{}
// get color health status
var beforeColor, afterColor col
var err error
switch i.Type {
case madmin.HealItemMetadata, madmin.HealItemBucket:
beforeColor, afterColor, err = getReplicatedFileHCCChange(i)
default:
if i.Type == madmin.HealItemObject {
itemStatus.Size = i.ObjectSize
}
beforeColor, afterColor, err = getObjectHCCChange(i)
}
if err != nil {
return err
}
itemStatus.Status = "success"
itemStatus.Before.Color = strings.ToLower(string(beforeColor))
itemStatus.After.Color = strings.ToLower(string(afterColor))
itemStatus.Type, itemStatus.Name = getHRITypeAndName(i)
itemStatus.Before.Online, itemStatus.After.Online = beforeUp, afterUp
itemStatus.Before.Missing, itemStatus.After.Missing = i.GetMissingCounts()
itemStatus.Before.Corrupted, itemStatus.After.Corrupted = i.GetCorruptedCounts()
itemStatus.Before.Offline, itemStatus.After.Offline = i.GetOfflineCounts()
itemStatus.Before.Drives = i.Before.Drives
itemStatus.After.Drives = i.After.Drives
h.ItemsHealthStatus = append(h.ItemsHealthStatus, itemStatus)
h.HealthBeforeCols[beforeColor]++
h.HealthAfterCols[afterColor]++
return nil
}
// getObjectHCCChange - returns before and after color change for
// objects
func getObjectHCCChange(h madmin.HealResultItem) (b, a col, err error) {
parityShards := h.ParityBlocks
dataShards := h.DataBlocks
onlineBefore, onlineAfter := h.GetOnlineCounts()
surplusShardsBeforeHeal := onlineBefore - dataShards
surplusShardsAfterHeal := onlineAfter - dataShards
b, err = getHColCode(surplusShardsBeforeHeal, parityShards)
if err != nil {
return
}
a, err = getHColCode(surplusShardsAfterHeal, parityShards)
return
}
// getReplicatedFileHCCChange - fetches health color code for metadata
// files that are replicated.
func getReplicatedFileHCCChange(h madmin.HealResultItem) (b, a col, err error) {
getColCode := func(numAvail int) (c col, err error) {
// calculate color code for replicated object similar
// to erasure coded objects
quorum := h.DiskCount/h.SetCount/2 + 1
surplus := numAvail/h.SetCount - quorum
parity := h.DiskCount/h.SetCount - quorum
c, err = getHColCode(surplus, parity)
return
}
onlineBefore, onlineAfter := h.GetOnlineCounts()
b, err = getColCode(onlineBefore)
if err != nil {
return
}
a, err = getColCode(onlineAfter)
return
}
func getHColCode(surplusShards, parityShards int) (c col, err error) {
if parityShards < 1 || parityShards > 8 || surplusShards > parityShards {
return c, fmt.Errorf("invalid parity shard count/surplus shard count given")
}
if surplusShards < 0 {
return colGrey, err
}
colRow := hColTable[parityShards]
for index, val := range colRow {
if val != -1 && surplusShards <= val {
return hColOrder[index], err
}
}
return c, fmt.Errorf("cannot get a heal color code")
}
func getHRITypeAndName(i madmin.HealResultItem) (typ, name string) {
name = fmt.Sprintf("%s/%s", i.Bucket, i.Object)
switch i.Type {
case madmin.HealItemMetadata:
typ = "system"
name = i.Detail
case madmin.HealItemBucketMetadata:
typ = "system"
name = "bucket-metadata:" + name
case madmin.HealItemBucket:
typ = "bucket"
case madmin.HealItemObject:
typ = "object"
default:
typ = fmt.Sprintf("!! Unknown heal result record %#v !!", i)
name = typ
}
return
}
// getHealOptionsFromReq return options from request for healing process
// path come as : `/heal/<namespace>/<tenantName>/bucket1`
// and query params come on request form
func getHealOptionsFromReq(req *http.Request) (*healOptions, error) {
hOptions := healOptions{}
re := regexp.MustCompile(`(/heal/)(.*?)(\?.*?$|$)`)
matches := re.FindAllSubmatch([]byte(req.URL.Path), -1)
// matches comes as e.g.
// [["...", "/heal/", "bucket1"]]
// [["/heal/" "/heal/" ""]]
if len(matches) == 0 || len(matches[0]) < 3 {
return nil, fmt.Errorf("invalid url: %s", req.URL.Path)
}
hOptions.BucketName = strings.TrimSpace(string(matches[0][2]))
hOptions.Prefix = req.FormValue("prefix")
hOptions.HealOpts.ScanMode = transformScanStr(req.FormValue("scan"))
if req.FormValue("force-start") != "" {
boolVal, err := strconv.ParseBool(req.FormValue("force-start"))
if err != nil {
return nil, err
}
hOptions.ForceStart = boolVal
}
if req.FormValue("force-stop") != "" {
boolVal, err := strconv.ParseBool(req.FormValue("force-stop"))
if err != nil {
return nil, err
}
hOptions.ForceStop = boolVal
}
// heal recursively
if req.FormValue("recursive") != "" {
boolVal, err := strconv.ParseBool(req.FormValue("recursive"))
if err != nil {
return nil, err
}
hOptions.HealOpts.Recursive = boolVal
}
// remove dangling objects in heal sequence
if req.FormValue("remove") != "" {
boolVal, err := strconv.ParseBool(req.FormValue("remove"))
if err != nil {
return nil, err
}
hOptions.HealOpts.Remove = boolVal
}
// only inspect data
if req.FormValue("dry-run") != "" {
boolVal, err := strconv.ParseBool(req.FormValue("dry-run"))
if err != nil {
return nil, err
}
hOptions.HealOpts.DryRun = boolVal
}
return &hOptions, nil
}
func transformScanStr(scanStr string) madmin.HealScanMode {
if scanStr == "deep" {
return madmin.HealDeepScan
}
return madmin.HealNormalScan
}

View File

@@ -1,270 +0,0 @@
// This file is part of MinIO Console Server
// Copyright (c) 2021 MinIO, Inc.
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package api
import (
"context"
"encoding/json"
"errors"
"net/http"
"net/url"
"testing"
"time"
"github.com/minio/madmin-go/v3"
"github.com/stretchr/testify/assert"
)
func TestHeal(t *testing.T) {
assert := assert.New(t)
client := AdminClientMock{}
mockWSConn := mockConn{}
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
function := "startHeal()"
mockResultItem1 := madmin.HealResultItem{
Type: madmin.HealItemObject,
SetCount: 1,
DiskCount: 4,
ParityBlocks: 2,
DataBlocks: 2,
Before: struct {
Drives []madmin.HealDriveInfo `json:"drives"`
}{
Drives: []madmin.HealDriveInfo{
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateMissing,
},
},
},
After: struct {
Drives []madmin.HealDriveInfo `json:"drives"`
}{
Drives: []madmin.HealDriveInfo{
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
},
},
}
mockResultItem2 := madmin.HealResultItem{
Type: madmin.HealItemBucket,
SetCount: 1,
DiskCount: 4,
ParityBlocks: 2,
DataBlocks: 2,
Before: struct {
Drives []madmin.HealDriveInfo `json:"drives"`
}{
Drives: []madmin.HealDriveInfo{
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateMissing,
},
},
},
After: struct {
Drives []madmin.HealDriveInfo `json:"drives"`
}{
Drives: []madmin.HealDriveInfo{
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
{
State: madmin.DriveStateOk,
},
},
},
}
mockHealTaskStatus := madmin.HealTaskStatus{
StartTime: time.Now().UTC().Truncate(time.Second * 2), // mock 2 sec duration
Items: []madmin.HealResultItem{
mockResultItem1,
mockResultItem2,
},
Summary: "finished",
}
testStreamSize := 1
testReceiver := make(chan healStatus, testStreamSize)
isClosed := false // testReceiver is closed?
testOptions := &healOptions{
BucketName: "testbucket",
Prefix: "",
ForceStart: false,
ForceStop: false,
}
// Test-1: startHeal send simple stream of data, no errors
minioHealMock = func(ctx context.Context, bucket, prefix string, healOpts madmin.HealOpts, clientToken string,
forceStart, forceStop bool,
) (healStart madmin.HealStartSuccess, healTaskStatus madmin.HealTaskStatus, err error) {
return healStart, mockHealTaskStatus, nil
}
writesCount := 1
// mock connection WriteMessage() no error
connWriteMessageMock = func(messageType int, data []byte) error {
// emulate that receiver gets the message written
var t healStatus
_ = json.Unmarshal(data, &t)
testReceiver <- t
if writesCount == testStreamSize {
// for testing we need to close the receiver channel
if !isClosed {
close(testReceiver)
isClosed = true
}
return nil
}
writesCount++
return nil
}
if err := startHeal(ctx, mockWSConn, client, testOptions); err != nil {
t.Errorf("Failed on %s:, error occurred: %s", function, err.Error())
}
// check that the TestReceiver got the same number of data from Console.
for i := range testReceiver {
assert.Equal(int64(1), i.ObjectsScanned)
assert.Equal(int64(1), i.ObjectsHealed)
assert.Equal(int64(2), i.ItemsScanned)
assert.Equal(int64(2), i.ItemsHealed)
assert.Equal(int64(0), i.HealthBeforeCols[colGreen])
assert.Equal(int64(1), i.HealthBeforeCols[colYellow])
assert.Equal(int64(1), i.HealthBeforeCols[colRed])
assert.Equal(int64(0), i.HealthBeforeCols[colGrey])
assert.Equal(int64(2), i.HealthAfterCols[colGreen])
assert.Equal(int64(0), i.HealthAfterCols[colYellow])
assert.Equal(int64(0), i.HealthAfterCols[colRed])
assert.Equal(int64(0), i.HealthAfterCols[colGrey])
}
// Test-2: startHeal error on init
minioHealMock = func(ctx context.Context, bucket, prefix string, healOpts madmin.HealOpts, clientToken string,
forceStart, forceStop bool,
) (healStart madmin.HealStartSuccess, healTaskStatus madmin.HealTaskStatus, err error) {
return healStart, mockHealTaskStatus, errors.New("error")
}
if err := startHeal(ctx, mockWSConn, client, testOptions); assert.Error(err) {
assert.Equal("error", err.Error())
}
// Test-3: getHealOptionsFromReq return heal options from request
u, _ := url.Parse("http://localhost/api/v1/heal/bucket1?prefix=file/&recursive=true&force-start=true&force-stop=true&remove=true&dry-run=true&scan=deep")
req := &http.Request{
URL: u,
}
opts, err := getHealOptionsFromReq(req)
if assert.NoError(err) {
expectedOptions := healOptions{
BucketName: "bucket1",
ForceStart: true,
ForceStop: true,
Prefix: "file/",
HealOpts: madmin.HealOpts{
Recursive: true,
DryRun: true,
ScanMode: madmin.HealDeepScan,
},
}
assert.Equal(expectedOptions.BucketName, opts.BucketName)
assert.Equal(expectedOptions.Prefix, opts.Prefix)
assert.Equal(expectedOptions.Recursive, opts.Recursive)
assert.Equal(expectedOptions.ForceStart, opts.ForceStart)
assert.Equal(expectedOptions.DryRun, opts.DryRun)
assert.Equal(expectedOptions.ScanMode, opts.ScanMode)
}
// Test-4: getHealOptionsFromReq return error if boolean value not valid
u, _ = url.Parse("http://localhost/api/v1/heal/bucket1?prefix=file/&recursive=nonbool&force-start=true&force-stop=true&remove=true&dry-run=true&scan=deep")
req = &http.Request{
URL: u,
}
_, err = getHealOptionsFromReq(req)
if assert.Error(err) {
assert.Equal("strconv.ParseBool: parsing \"nonbool\": invalid syntax", err.Error())
}
// Test-5: getHealOptionsFromReq return error if boolean value not valid
u, _ = url.Parse("http://localhost/api/v1/heal/bucket1?prefix=file/&recursive=true&force-start=true&force-stop=true&remove=nonbool&dry-run=true&scan=deep")
req = &http.Request{
URL: u,
}
_, err = getHealOptionsFromReq(req)
if assert.Error(err) {
assert.Equal("strconv.ParseBool: parsing \"nonbool\": invalid syntax", err.Error())
}
// Test-6: getHealOptionsFromReq return error if boolean value not valid
u, _ = url.Parse("http://localhost/api/v1/heal/bucket1?prefix=file/&recursive=true&force-start=nonbool&force-stop=true&remove=true&dry-run=true&scan=deep")
req = &http.Request{
URL: u,
}
_, err = getHealOptionsFromReq(req)
if assert.Error(err) {
assert.Equal("strconv.ParseBool: parsing \"nonbool\": invalid syntax", err.Error())
}
// Test-7: getHealOptionsFromReq return error if boolean value not valid
u, _ = url.Parse("http://localhost/api/v1/heal/bucket1?prefix=file/&recursive=true&force-start=true&force-stop=nonbool&remove=true&dry-run=true&scan=deep")
req = &http.Request{
URL: u,
}
_, err = getHealOptionsFromReq(req)
if assert.Error(err) {
assert.Equal("strconv.ParseBool: parsing \"nonbool\": invalid syntax", err.Error())
}
// Test-8: getHealOptionsFromReq return error if boolean value not valid
u, _ = url.Parse("http://localhost/api/v1/heal/bucket1?prefix=file/&recursive=true&force-start=true&force-stop=true&remove=true&dry-run=nonbool&scan=deep")
req = &http.Request{
URL: u,
}
_, err = getHealOptionsFromReq(req)
if assert.Error(err) {
assert.Equal("strconv.ParseBool: parsing \"nonbool\": invalid syntax", err.Error())
}
}

View File

@@ -63,7 +63,6 @@ type wsAdminClient struct {
// ConsoleWebsocket interface of a Websocket Client
type ConsoleWebsocket interface {
watch(options watchOptions)
heal(opts healOptions)
}
type wsS3Client struct {
@@ -245,20 +244,6 @@ func serveWS(w http.ResponseWriter, req *http.Request) {
return
}
go wsAdminClient.healthInfo(ctx, deadline)
case strings.HasPrefix(wsPath, `/heal`):
hOptions, err := getHealOptionsFromReq(req)
if err != nil {
ErrorWithContext(ctx, fmt.Errorf("error getting heal options: %v", err))
closeWsConn(conn)
return
}
wsAdminClient, err := newWebSocketAdminClient(conn, session)
if err != nil {
ErrorWithContext(ctx, err)
closeWsConn(conn)
return
}
go wsAdminClient.heal(ctx, hOptions)
case strings.HasPrefix(wsPath, `/watch`):
wOptions, err := getWatchOptionsFromReq(req)
if err != nil {
@@ -506,21 +491,6 @@ func (wsc *wsS3Client) watch(ctx context.Context, params *watchOptions) {
sendWsCloseMessage(wsc.conn, err)
}
func (wsc *wsAdminClient) heal(ctx context.Context, opts *healOptions) {
defer func() {
LogInfo("heal stopped")
// close connection after return
wsc.conn.close()
}()
LogInfo("heal started")
ctx = wsReadClientCtx(ctx, wsc.conn)
err := startHeal(ctx, wsc.conn, wsc.client, opts)
sendWsCloseMessage(wsc.conn, err)
}
func (wsc *wsAdminClient) healthInfo(ctx context.Context, deadline *time.Duration) {
defer func() {
LogInfo("health info stopped")

View File

@@ -1,16 +0,0 @@
{
"Version": "2012-10-17",
"Statement": [
{
"Action": ["admin:Heal"],
"Effect": "Allow",
"Sid": ""
},
{
"Action": ["s3:ListBucket"],
"Effect": "Allow",
"Resource": ["arn:aws:s3:::*"],
"Sid": ""
}
]
}

View File

@@ -30,7 +30,6 @@ create_policies() {
mc admin policy create minio dashboard-$TIMESTAMP web-app/tests/policies/dashboard.json
mc admin policy create minio diagnostics-$TIMESTAMP web-app/tests/policies/diagnostics.json
mc admin policy create minio groups-$TIMESTAMP web-app/tests/policies/groups.json
mc admin policy create minio heal-$TIMESTAMP web-app/tests/policies/heal.json
mc admin policy create minio iampolicies-$TIMESTAMP web-app/tests/policies/iamPolicies.json
mc admin policy create minio logs-$TIMESTAMP web-app/tests/policies/logs.json
mc admin policy create minio notificationendpoints-$TIMESTAMP web-app/tests/policies/notificationEndpoints.json