livrare lot 2

This commit is contained in:
EVOTECH IT SRL 2026-07-10 03:39:53 -07:00
commit 8ecc78e729
763 changed files with 164593 additions and 0 deletions

View file

@ -0,0 +1,104 @@
# DIDI PostgreSQL HA — Patroni + etcd + HAProxy (IaC livrabil)
Pachet **reproductibil** care livrează clusterul HA PostgreSQL al platformei DiDi
ca Infrastructure-as-Code. Aceeași arhitectură rulează în producție pe VM-uri
dedicate (vezi `../MIGRATION.md`); acest compose o reproduce integral pe un
singur host pentru demo, recepție, DR-rehearsal și medii de test.
## Arhitectură
```
┌────────────────────┐
apps ──5000──▶ │ HAProxy │ ──▶ /primary (Patroni REST :8008)
apps ──5001──▶ │ (LB + healthcheck)│ ──▶ /replica
└─────────┬──────────┘
┌───────────────┼───────────────┐
┌─────▼─────┐ ┌─────▼─────┐ ┌─────▼─────┐
│ pg-node1 │ │ pg-node2 │ │ pg-node3 │ Spilo = PostgreSQL 16
│ Patroni │ │ Patroni │ │ Patroni │ + Patroni (Zalando)
└─────┬─────┘ └─────┬─────┘ └─────┬─────┘
└───────────────┼───────────────┘
┌────────▼────────┐
│ etcd1/2/3 (DCS) │ quorum leader-election
└─────────────────┘
```
| Rol | Producție (VM-uri) | Acest pachet |
|---|---|---|
| PG + Patroni ×3 | 10.11.50.160162 | `pg-node1..3` (Spilo 16) |
| etcd quorum ×3 | 10.11.50.163165 | `etcd1..3` (v3.5) |
| HAProxy | 10.11.50.166 + 169 (VIP .167) | `haproxy` :5000/:5001 |
| Backup | pgBackRest (10.11.50.168, NFS) | vezi §Backup |
## Pornire
```bash
docker compose up -d
# election durează ~30-60s; verifică:
docker exec didi-ha-pg1 patronictl list
```
Conectare (contract identic cu producția):
```bash
PGPASSWORD=didi-super-secret psql -h localhost -p 5000 -U postgres # RW (leader)
PGPASSWORD=didi-super-secret psql -h localhost -p 5001 -U postgres # RO (replici)
```
Restaurare schema DIDI (bos_parammgmt / bos_analysis / bos_sysadmin / bos_subscriber):
```bash
PGPASSWORD=didi-super-secret psql -h localhost -p 5000 -U postgres \
-f ../DIDI_full_export_2026-07-02.sql
```
## Test failover (drill de recepție)
```bash
# 1. află liderul
docker exec didi-ha-pg1 patronictl list
# 2. omoară-l
docker stop didi-ha-pg2 # (dacă pg2 e leader)
# 3. Patroni promovează o replică în secunde; HAProxy reroutează :5000
# automat (healthcheck /primary la 3s, fall 3). Aplicațiile nu schimbă
# nimic — se reconectează pe același endpoint.
docker exec didi-ha-pg1 patronictl list
# 4. reintră nodul căzut ca replică:
docker start didi-ha-pg2
```
Switchover planificat (fără downtime):
```bash
docker exec didi-ha-pg1 patronictl switchover didi --force
```
## Parametri
| Env | Default | Rol |
|---|---|---|
| `PG_SUPERUSER_PASSWORD` | `didi-super-secret` | postgres superuser |
| `PG_ADMIN_PASSWORD` | `didi-admin-secret` | admin role |
| `PG_STANDBY_PASSWORD` | `didi-standby-secret` | replicare streaming |
**Schimbă-le obligatoriu în producție** (`.env` lângă compose).
## Backup
În producție backup-ul e pgBackRest (full zilnic + WAL archiving pe NFS,
nod dedicat). Pe acest pachet, echivalentul minim:
```bash
docker exec didi-ha-pg1 su postgres -c \
'pg_basebackup -h localhost -p 5432 -D /tmp/didi-backup -Ft -z -Xs'
```
## Relația cu livrabilul Lot 2
- Modulul 5 (Baze de date SQL) cere PostgreSQL cu HA; oferta specifică
Patroni + HAProxy. Acest director este implementarea IaC livrată —
reproductibilă pe orice host Docker, plus instanțierea de producție
documentată în `MIGRATION.md`.
- Aplicațiile (agent-v3, didiFramework) sunt agnostice: `PG_HOST:PG_PORT`
arată fie spre VIP-ul de producție (`10.11.50.167:5000`), fie spre acest
cluster local (`localhost:5000`) — același contract, zero modificări de cod.

View file

@ -0,0 +1,116 @@
# ============================================================================
# DIDI PostgreSQL HA cluster — Patroni + etcd + HAProxy (IaC, reproducible)
#
# Containerized mirror of the production topology (see ../MIGRATION.md):
# prod: 3× PG/Patroni (10.11.50.160-162) + 3× etcd (163-165)
# + 2× HAProxy (166/169, VIP 167) + pgBackRest (168)
# here: 3× Spilo (Patroni+PG, Zalando) + 3× etcd + 1× HAProxy
# → same failover semantics, single-host footprint for
# demo/recepție/DR-rehearsal.
#
# Endpoints (identical contract to production):
# localhost:5000 → leader (read-write) [HAProxy checks Patroni /primary]
# localhost:5001 → replicas (read-only) [HAProxy checks Patroni /replica]
# localhost:7000 → HAProxy stats UI
#
# Usage:
# docker compose up -d
# # wait ~30s for leader election, then:
# psql -h localhost -p 5000 -U postgres # password: $PG_SUPERUSER_PASSWORD
# # restore DIDI schema:
# psql -h localhost -p 5000 -U postgres -f ../DIDI_full_export_2026-07-02.sql
# # failover drill:
# docker compose stop $(docker compose ps --format '{{.Name}}' | head -1)
# # → a replica is promoted in seconds; :5000 keeps serving writes.
# ============================================================================
x-etcd-common: &etcd-common
image: quay.io/coreos/etcd:v3.5.16
restart: unless-stopped
networks: [didi-ha]
environment: &etcd-env
ETCD_INITIAL_CLUSTER: etcd1=http://etcd1:2380,etcd2=http://etcd2:2380,etcd3=http://etcd3:2380
ETCD_INITIAL_CLUSTER_STATE: new
ETCD_INITIAL_CLUSTER_TOKEN: didi-pg-ha
ETCD_AUTO_COMPACTION_RETENTION: "1"
ETCD_ENABLE_V2: "true"
x-spilo-common: &spilo-common
image: ghcr.io/zalando/spilo-16:3.3-p3
restart: unless-stopped
networks: [didi-ha]
environment: &spilo-env
SCOPE: didi # Patroni cluster name (etcd namespace)
PGVERSION: "16"
ETCD3_HOSTS: "'etcd1:2379','etcd2:2379','etcd3:2379'"
PGPASSWORD_SUPERUSER: ${PG_SUPERUSER_PASSWORD:-didi-super-secret}
PGPASSWORD_ADMIN: ${PG_ADMIN_PASSWORD:-didi-admin-secret}
PGPASSWORD_STANDBY: ${PG_STANDBY_PASSWORD:-didi-standby-secret}
ALLOW_NOSSL: "true"
healthcheck:
test: ["CMD-SHELL", "curl -sf http://localhost:8008/health || exit 1"]
interval: 10s
timeout: 5s
retries: 6
start_period: 60s
services:
etcd1:
<<: *etcd-common
container_name: didi-ha-etcd1
command: etcd --name etcd1
--listen-peer-urls http://0.0.0.0:2380 --initial-advertise-peer-urls http://etcd1:2380
--listen-client-urls http://0.0.0.0:2379 --advertise-client-urls http://etcd1:2379
etcd2:
<<: *etcd-common
container_name: didi-ha-etcd2
command: etcd --name etcd2
--listen-peer-urls http://0.0.0.0:2380 --initial-advertise-peer-urls http://etcd2:2380
--listen-client-urls http://0.0.0.0:2379 --advertise-client-urls http://etcd2:2379
etcd3:
<<: *etcd-common
container_name: didi-ha-etcd3
command: etcd --name etcd3
--listen-peer-urls http://0.0.0.0:2380 --initial-advertise-peer-urls http://etcd3:2380
--listen-client-urls http://0.0.0.0:2379 --advertise-client-urls http://etcd3:2379
pg-node1:
<<: *spilo-common
container_name: didi-ha-pg1
hostname: pg-node1
depends_on: [etcd1, etcd2, etcd3]
volumes: [pg1-data:/home/postgres/pgdata]
pg-node2:
<<: *spilo-common
container_name: didi-ha-pg2
hostname: pg-node2
depends_on: [etcd1, etcd2, etcd3]
volumes: [pg2-data:/home/postgres/pgdata]
pg-node3:
<<: *spilo-common
container_name: didi-ha-pg3
hostname: pg-node3
depends_on: [etcd1, etcd2, etcd3]
volumes: [pg3-data:/home/postgres/pgdata]
haproxy:
image: haproxy:2.9-alpine
container_name: didi-ha-haproxy
restart: unless-stopped
networks: [didi-ha]
depends_on: [pg-node1, pg-node2, pg-node3]
ports:
- "5000:5000" # read-write → Patroni leader
- "5001:5001" # read-only → replicas
- "7000:7000" # stats UI
volumes:
- ./haproxy/haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg:ro
volumes:
pg1-data:
pg2-data:
pg3-data:
networks:
didi-ha:
name: didi-ha

View file

@ -0,0 +1,46 @@
# HAProxy for DIDI PostgreSQL HA — routes by Patroni REST health checks.
# Mirrors the production LB config (10.11.50.166/169 → VIP 167).
#
# :5000 → the ONE node whose Patroni answers 200 on /primary (leader, RW)
# :5001 → nodes answering 200 on /replica (round-robin, RO)
#
# On failover Patroni flips the health endpoints; HAProxy reroutes in
# (inter × fall) ≈ 9s worst case without client config changes.
global
maxconn 300
log stdout format raw local0
defaults
log global
mode tcp
retries 2
timeout client 30m
timeout connect 4s
timeout server 30m
timeout check 5s
listen stats
mode http
bind *:7000
stats enable
stats uri /
listen postgres_write
bind *:5000
option httpchk GET /primary
http-check expect status 200
default-server inter 3s fall 3 rise 2 on-marked-down shutdown-sessions
server pg-node1 pg-node1:5432 check port 8008
server pg-node2 pg-node2:5432 check port 8008
server pg-node3 pg-node3:5432 check port 8008
listen postgres_read
bind *:5001
balance roundrobin
option httpchk GET /replica
http-check expect status 200
default-server inter 3s fall 3 rise 2 on-marked-down shutdown-sessions
server pg-node1 pg-node1:5432 check port 8008
server pg-node2 pg-node2:5432 check port 8008
server pg-node3 pg-node3:5432 check port 8008