ops: add production observability and recovery

This commit is contained in:
hailin
2026-07-27 19:06:34 +08:00
parent e69368a2f3
commit b02d18d3b9
33 changed files with 8832 additions and 1 deletions

View File

@@ -0,0 +1,194 @@
[CmdletBinding()]
param(
[string]$ComposeFile = "",
[switch]$SkipBuild
)
$ErrorActionPreference = "Stop"
$scriptDirectory = Split-Path -Parent $MyInvocation.MyCommand.Path
$infraDirectory = [IO.Path]::GetFullPath((Join-Path $scriptDirectory ".."))
$projectDirectory = [IO.Path]::GetFullPath((Join-Path $infraDirectory ".."))
if (-not $ComposeFile) {
$ComposeFile = Join-Path $infraDirectory "docker-compose.prod.yml"
}
$ComposeFile = (Resolve-Path -LiteralPath $ComposeFile).Path
$projectName = "ai-ticket-restore-drill-$PID"
$databasePassword = "restore-drill-postgres-$PID"
$minioUser = "restore-drill-minio"
$minioPassword = "restore-drill-minio-$PID"
$backupRoot = Join-Path $projectDirectory "backups\$projectName"
$stopwatch = [Diagnostics.Stopwatch]::StartNew()
$savedEnvironment = @{
COMPOSE_PROJECT_NAME = $env:COMPOSE_PROJECT_NAME
POSTGRES_DB = $env:POSTGRES_DB
POSTGRES_USER = $env:POSTGRES_USER
POSTGRES_PASSWORD = $env:POSTGRES_PASSWORD
MINIO_ROOT_USER = $env:MINIO_ROOT_USER
MINIO_ROOT_PASSWORD = $env:MINIO_ROOT_PASSWORD
}
$env:COMPOSE_PROJECT_NAME = $projectName
$env:POSTGRES_DB = "ai_ticket_platform"
$env:POSTGRES_USER = "ai_ticket"
$env:POSTGRES_PASSWORD = $databasePassword
$env:MINIO_ROOT_USER = $minioUser
$env:MINIO_ROOT_PASSWORD = $minioPassword
function Invoke-Docker {
param([Parameter(Mandatory = $true)][string[]]$Arguments)
& docker @Arguments
if ($LASTEXITCODE -ne 0) {
throw "Docker 执行失败:docker $($Arguments -join ' ')"
}
}
function Invoke-DjangoCode {
param(
[Parameter(Mandatory = $true)][string[]]$RunPrefix,
[Parameter(Mandatory = $true)][string]$Code,
[Parameter(Mandatory = $true)][string]$FailureMessage
)
$bytes = [Text.Encoding]::UTF8.GetBytes($Code)
$runner = "exec(bytes([" + ($bytes -join ",") + "]))"
& docker @RunPrefix python manage.py shell -c $runner
if ($LASTEXITCODE -ne 0) {
throw $FailureMessage
}
}
$backendEnvironment = @(
"-e", "POSTGRES_HOST=postgres",
"-e", "POSTGRES_DB=ai_ticket_platform",
"-e", "POSTGRES_USER=ai_ticket",
"-e", "POSTGRES_PASSWORD=$databasePassword",
"-e", "MINIO_ENDPOINT=minio:9000",
"-e", "MINIO_ROOT_USER=$minioUser",
"-e", "MINIO_ROOT_PASSWORD=$minioPassword",
"-e", "CELERY_BROKER_URL=redis://redis:6379/0",
"-e", "CELERY_RESULT_BACKEND=redis://redis:6379/1"
)
$seedCode = @'
from django.contrib.auth import get_user_model
from apps.tickets.models import Ticket, TicketAttachment
user, _ = get_user_model().objects.get_or_create(
username="restore_drill",
defaults={"email": "restore-drill@example.invalid"},
)
ticket, _ = Ticket.objects.update_or_create(
ticket_no="RESTORE-DRILL-001",
defaults={
"title": "隔离恢复演练工单",
"description": "restore-drill-marker",
"requester": user,
"created_by": user,
},
)
TicketAttachment.objects.update_or_create(
ticket=ticket,
storage_path="restore-drill/marker.txt",
defaults={
"file_name": "marker.txt",
"mime_type": "text/plain",
"file_size": 20,
"created_by": user,
},
)
print(f"seeded_ticket={ticket.ticket_no}")
'@
$deleteCode = @'
from apps.tickets.models import Ticket
Ticket.objects.filter(ticket_no="RESTORE-DRILL-001").delete()
print("seed_deleted=true")
'@
$verifyCode = @'
from apps.tickets.models import Ticket
ticket = Ticket.objects.get(ticket_no="RESTORE-DRILL-001")
attachment = ticket.attachments.get(storage_path="restore-drill/marker.txt")
assert ticket.description == "restore-drill-marker"
assert attachment.file_name == "marker.txt"
print(f"restored_ticket={ticket.ticket_no}")
print(f"restored_attachment={attachment.storage_path}")
'@
try {
Invoke-Docker @("compose", "-f", $ComposeFile, "up", "-d", "--wait", "postgres", "minio")
if (-not $SkipBuild) {
Invoke-Docker @("compose", "-f", $ComposeFile, "build", "backend")
}
$runPrefix = @(
"compose", "-f", $ComposeFile,
"run", "--rm", "-T", "--no-deps"
) + $backendEnvironment + @("backend")
Invoke-Docker ($runPrefix + @("python", "manage.py", "migrate", "--noinput"))
Invoke-DjangoCode -RunPrefix $runPrefix -Code $seedCode -FailureMessage "演练数据创建失败"
Invoke-Docker @(
"compose", "-f", $ComposeFile, "exec", "-T", "minio",
"sh", "-c",
'mc alias set drill http://127.0.0.1:9000 "$MINIO_ROOT_USER" "$MINIO_ROOT_PASSWORD" >/dev/null && mc mb --ignore-existing drill/ai-ticket-platform >/dev/null && printf restore-drill-marker | mc pipe drill/ai-ticket-platform/restore-drill/marker.txt >/dev/null'
)
& (Join-Path $scriptDirectory "backup.ps1") `
-ComposeFile $ComposeFile `
-BackupRoot $backupRoot `
-PostgresDatabase "ai_ticket_platform" `
-PostgresUser "ai_ticket" `
-SkipApplicationServices
$backupDirectory = Get-ChildItem -LiteralPath $backupRoot -Directory |
Sort-Object Name -Descending |
Select-Object -First 1
if (-not $backupDirectory) { throw "未找到演练备份目录" }
Invoke-DjangoCode -RunPrefix $runPrefix -Code $deleteCode -FailureMessage "演练数据破坏步骤失败"
Invoke-Docker @(
"compose", "-f", $ComposeFile, "exec", "-T", "minio",
"mc", "rm", "--force", "drill/ai-ticket-platform/restore-drill/marker.txt"
)
& (Join-Path $scriptDirectory "restore.ps1") `
-BackupDirectory $backupDirectory.FullName `
-ComposeFile $ComposeFile `
-PostgresDatabase "ai_ticket_platform" `
-PostgresUser "ai_ticket" `
-Confirm "restore-ai-ticket-platform" `
-SkipApplicationServices
Invoke-DjangoCode -RunPrefix $runPrefix -Code $verifyCode -FailureMessage "数据库业务冒烟失败"
$objectMarker = & docker compose -f $ComposeFile exec -T minio `
mc cat drill/ai-ticket-platform/restore-drill/marker.txt
if ($LASTEXITCODE -ne 0 -or "$objectMarker".Trim() -ne "restore-drill-marker") {
throw "对象存储业务冒烟失败"
}
Invoke-Docker ($runPrefix + @("python", "manage.py", "check"))
$stopwatch.Stop()
[pscustomobject]@{
status = "passed"
compose_project = $projectName
backup_id = $backupDirectory.Name
duration_seconds = [Math]::Round($stopwatch.Elapsed.TotalSeconds, 2)
database_smoke = "ticket_and_attachment_restored"
object_storage_smoke = "marker_restored"
django_check = "passed"
backup_directory = $backupDirectory.FullName
} | ConvertTo-Json
}
finally {
& docker compose -f $ComposeFile down -v --remove-orphans --rmi local
foreach ($entry in $savedEnvironment.GetEnumerator()) {
if ($null -eq $entry.Value) {
Remove-Item "Env:$($entry.Key)" -ErrorAction SilentlyContinue
}
else {
Set-Item "Env:$($entry.Key)" $entry.Value
}
}
}