railiance-platform/scripts/verify_scaleway_primary_restore.py

#!/usr/bin/env python3
"""Bounded shared/Forgejo database recovery from the existing Scaleway primary; no in-place restore."""
import argparse
import copy
from datetime import datetime, timezone
import json
import os
from pathlib import Path
import secrets
import subprocess
import time
from state_hub_preflight_lane import assert_cluster


def run(k, receipt, source_cluster="apps-pg"):
    if source_cluster not in ("apps-pg", "forgejo-db"): raise ValueError("unsupported source cluster")
    def cmd(args,payload=None,check=True):
        r=subprocess.run(k+args,input=None if payload is None else json.dumps(payload).encode(),capture_output=True,timeout=180)
        if check and r.returncode: raise RuntimeError('bounded_restore_command_failed')
        return r
    def get(args): return json.loads(cmd(args).stdout)
    assert_cluster(k)
    live=get(['-n','databases','get','cluster',source_cluster,'-o','json'])
    store=copy.deepcopy(live['spec']['backup']['barmanObjectStore'])
    if store['destinationPath']!='s3://railiance-platform-pg-backup/platform-pg/'+source_cluster+'/' or store['endpointURL']!='https://s3.nl-ams.scw.cloud': raise RuntimeError('unexpected_primary_destination')
    recent=datetime.fromisoformat(live['status']['lastSuccessfulBackup'].replace('Z','+00:00'))
    if not 0 <= (datetime.now(timezone.utc)-recent).total_seconds() < 129600: raise RuntimeError('primary_backup_not_fresh')
    name=source_cluster+'-recovery-'+secrets.token_hex(4)
    receipt.update(namespace=name,primary_destination=store['destinationPath'],last_successful_backup=live['status']['lastSuccessfulBackup'],source='Scaleway Barman base backup and WAL',started_at=datetime.now(timezone.utc).isoformat())
    created=False
    try:
        cmd(['create','namespace',name]);created=True
        receipt['stage']='credential_projection'
        # Exact existing S3 fields only; capture JSON in memory and never print it.
        for secret_name in {store['s3Credentials'][key]['name'] for key in ('accessKeyId','secretAccessKey')}:
            source=get(['-n','databases','get','secret',secret_name,'-o','json'])
            fields={store['s3Credentials'][key]['key'] for key in ('accessKeyId','secretAccessKey') if store['s3Credentials'][key]['name']==secret_name}
            doc={'apiVersion':'v1','kind':'Secret','metadata':{'name':secret_name,'namespace':name},'type':'Opaque','data':{f:source['data'][f] for f in fields}}
            cmd(['create','-f','-'],doc)
        policy={'apiVersion':'networking.k8s.io/v1','kind':'NetworkPolicy','metadata':{'name':'isolated-recovery','namespace':name},'spec':{'podSelector':{},'policyTypes':['Ingress','Egress'],'ingress':[{'from':[{'namespaceSelector':{'matchLabels':{'kubernetes.io/metadata.name':'cnpg-system'}}}],'ports':[{'port':8000,'protocol':'TCP'}]}],'egress':[{'ports':[{'port':p,'protocol':proto} for p,proto in [(443,'TCP'),(6443,'TCP'),(53,'TCP'),(53,'UDP')]]}]}}
        cmd(['create','-f','-'],policy)
        store['serverName']=source_cluster
        cluster={'apiVersion':'postgresql.cnpg.io/v1','kind':'Cluster','metadata':{'name':name,'namespace':name},'spec':{'instances':1,'imageName':live['spec']['imageName'],'enableSuperuserAccess':False,'storage':{'size':'10Gi'},'resources':{'requests':{'cpu':'100m','memory':'256Mi'},'limits':{'cpu':'1','memory':'1Gi'}},'bootstrap':{'recovery':{'source':source_cluster}},'externalClusters':[{'name':source_cluster,'barmanObjectStore':store}]}}
        receipt['stage']='physical_restore'
        cmd(['create','--dry-run=server','-f','-'],cluster)
        started=time.monotonic();cmd(['create','-f','-'],cluster)
        for _ in range(60):
            state=get(['-n',name,'get','cluster',name,'-o','json'])
            if any(c['type']=='Ready' and c['status']=='True' for c in state.get('status',{}).get('conditions',[])): break
            time.sleep(5)
        else: raise RuntimeError('scratch_restore_not_ready')
        receipt['restore_seconds']=round(time.monotonic()-started,2)
        pod=state['status']['currentPrimary']
        def sql(database,query): return cmd(['-n',name,'exec',pod,'-c','postgres','--','psql','-U','postgres','-d',database,'-Atqc',query]).stdout.decode().strip()
        receipt['stage']='database_acceptance'
        databases=sql('postgres',"SELECT datname FROM pg_database WHERE NOT datistemplate ORDER BY datname;").splitlines()
        receipt['databases']=databases
        if source_cluster=='apps-pg':
            if not {'apps_meta','coulomb_social_db','vergabe_db'}.issubset(databases): raise RuntimeError('restored_database_missing')
            receipt['public_table_counts']={db:int(sql(db,"SELECT count(*) FROM pg_tables WHERE schemaname='public';")) for db in ['coulomb_social_db','vergabe_db']}
            roles=sql('postgres',"SELECT rolname || ':' || rolconnlimit FROM pg_roles WHERE rolname IN ('coulomb_social','vergabe') ORDER BY rolname;").splitlines()
            if roles!=['coulomb_social:20','vergabe:20']: raise RuntimeError('restored_role_boundary_mismatch')
            receipt['consumer_connection_limits_preserved']=True
        else:
            if 'forgejo' not in databases: raise RuntimeError('restored_database_missing')
            receipt['repository_count']=int(sql('forgejo','SELECT count(*) FROM repository;'))
            receipt['user_count']=int(sql('forgejo','SELECT count(*) FROM "user";'))
            receipt['package_blob_count']=int(sql('forgejo','SELECT count(*) FROM package_blob;'))
            if min(receipt['repository_count'],receipt['user_count'])<=0: raise RuntimeError('restored_forgejo_metadata_empty')
        live_after=get(['-n','databases','get','cluster',source_cluster,'-o','json'])
        receipt['production_ready']=any(c['type']=='Ready' and c['status']=='True' for c in live_after['status']['conditions'])
        if not receipt['production_ready']: raise RuntimeError('production_not_ready')
        receipt['status']='verified'
    finally:
        if created:
            receipt['cleanup']=cmd(['delete','namespace',name,'--wait=true','--timeout=120s'],check=False).returncode==0
        receipt['finished_at']=datetime.now(timezone.utc).isoformat()


def main():
    p=argparse.ArgumentParser(description=__doc__);p.add_argument('--kubeconfig',required=True);p.add_argument('--receipt',required=True);p.add_argument('--source-cluster',choices=['apps-pg','forgejo-db'],default='apps-pg');a=p.parse_args()
    receipt={'schema':'platform.scaleway-primary-restore.v1','status':'failed'}
    fd=os.open(a.receipt,os.O_WRONLY|os.O_CREAT|os.O_EXCL,0o600)
    try: run(['kubectl','--kubeconfig',a.kubeconfig],receipt,a.source_cluster)
    except Exception: receipt['error']='primary_restore_failed'
    finally:
        with os.fdopen(fd,'w') as f: json.dump(receipt,f,indent=2)
    return int(receipt['status']!='verified' or not receipt.get('cleanup'))
if __name__=='__main__': raise SystemExit(main())