Commit Graph

5 Commits

Author SHA1 Message Date
chrisfu
5618b662dd Remove prole-db-manager; simplify deployment via prole-authority; fix pg18 downgrade & cluster name
Summary:
Removed the prole-db-manager microservice and simplified deployment to use
prole-authority as the internal management and authorization point. Fixed two
blocking bugs that prevented silent install from completing on knoe-dev-cluster.

Removed: prole-db-manager
- Deleted db-manager-deployment.yaml and db-manager-service.yaml from opentofu manifests
- Deleted src/db-manager/ (Dockerfile, server.js, package.json, tests)
- Removed prole-db-manager port-forward mapping from installer/core/env.py
- Removed init_db_manager.sh from Initialization Scripts (milestones.py, actions.py)
- Removed init_certmgr.sh and init_db_manager.sh tabs from services screen (services.py)
- Removed live k8s Deployment/Service from knoe-dev-cluster

Fixed: PostgreSQL version downgrade error (pg17 -> pg18)
- Created conf/postgresql/.version with value 18
- Updated k8s/prole/prole-db.yaml and prole-db-recovery.yaml.tpl imageName to prole-db:18-089
- Fixed _init_database_options_state() to restore saved version_type from prole.cfg
  so db_version_type defaults to v18 (pg18) instead of silently reverting to pg17
- Added database_options.* keys to _collect_input_snapshot() in cfg.py so
  distribution, version_type, and all extension toggles persist to prole.cfg

Fixed: Cluster name inconsistency
- Removed stale prole-dev-cluster references; all scripts now use knoe-dev-cluster
- Added knoe-dev-cluster to mode-detection case in etc/prole_cfg.sh

Config: conf/prole.cfg
- Set kerberos_config.enabled = False, KERBEROS_AUTO_ENABLED = False
- Added database_options.distribution = percona, version_type = v18
- Added all 13 extension flags set to True (postgis, pgvector, pgcrypto, pgaudit,
  pg_repack, pg_stat_statements, pg_buffercache, pg_freespacemap, pgrowlocks,
  postgres_fdw, dblink, pg_stat_monitor, pgbadger)

Verification:
./install.py -s -l -v -c conf/prole.cfg completed successfully.
CNPG deployed prole-db:18-089 to knoe-dev-cluster; all milestones passed.

Co-authored-by: Junie <junie@jetbrains.com>
2026-03-01 20:40:44 -08:00
chrisfu
765aa0b926 Checkpoint: cluster env UI layout + k3s common services
- Tighten Cluster Environment screen layout; switch Service/Prod to kubectx context selection; keep namespace and key controls on one line; ensure Repair button remains reachable.

- Add UI layout regression test to render with large mock data and assert key widgets remain visible and console is scrollable.

- Make kubeconfig generation deterministic under tests by avoiding overwriting cert-based kubeconfigs; write token sidecar kubeconfig when needed.

- Update common-services init scripts and add k3s/Helm deployment bits (svc-check, Kong/CertMgr tasks).
2026-02-27 14:35:14 -08:00
chrisfu
dafed53810 Enhance build system with 'make test' and fix installer test debris. Added 'make test' to Makefile with 'pyconv' (black) integration and coverage summary. Reformatted codebase with black. Fixed 'install.py' test bug where MagicMock objects created directory debris by improving mocking and patching in 'tests/test_navigation.py', 'tests/test_service_layer_navigation.py', and 'tests/test_install_logic.py'. 2026-02-25 14:55:05 -08:00
chrisfu
b03efa8f69 Kong API gateway, docker-import preload, OpenTofu graceful fallback, milestone fix
Kong API Gateway (replacing prole nginx):
- Add etc/init_kong.sh provisioning script (DB-less mode, prole-db namespace)
- Add kong-deployment.yaml and kong-service.yaml manifests
- Rewire ingress rules (svc/git/api.prole.org) to prole-db-kong:8000
- Update kustomization.yaml to reference kong manifests

PostgREST & DB Manager in prole-db namespace:
- Add etc/init_postgrest.sh and etc/init_db_manager.sh scripts
- Add postgrest/db-manager deployment and service manifests
- Add src/db-manager/ Node.js REST endpoint for backup triggers
- Default NAMESPACE changed to prole-db in both scripts

Docker image pre-load from PROLE_DATA/docker-import:
- Add _preload_docker_images() to init_common_services.sh
- Scan for .tar files exported by final_deployment.sh
- Import via k3d image import (k3d) or ctr (k3s) before deployments
- Increase rollout timeouts to 300s (configurable via ROLLOUT_TIMEOUT) in init_openbao.sh, init_opentofu.sh, init_garage_store.sh, init_registry.sh

OpenTofu password resolution fix:
- Add Kubernetes secret fallback in resolve_admin_password()
- Change hard exit 1 to graceful return 1 with warning
- Wrap call in if-guard so set -e doesn't abort the script chain

Milestone fix (init scripts not running):
- Add init_kong.sh, init_postgrest.sh, init_db_manager.sh to InitializationScriptsMilestone.execute() script list and arg branches
- Previously only actions.py had these; milestones.py was missing them

Installer integration:
- Add Kong/PostgREST/DB Manager to silent installer _step_init_scripts
- Add corresponding tabs and execution blocks in UI services.py
2026-02-22 00:57:49 -08:00
chrisfu
ec50f82cc8 refactor: deduplicate installer business logic & split screens.py into package
Separation of concerns: merge silent/UI actions & modularize screens.

ProleInstallerBase (actions.py): Created shared base class with 37 deduplicated methods previously duplicated between ProleSilentInstaller and ProleInstaller. Namespace, environment, secret, deployment, port-forward, authority/repair, image, and logging helpers now defined once. Subclasses override _get_input() to bridge their data-access layers.

screens.py -> screens/ package (18 mixin modules): Split 10,234-line monolithic screens.py into focused mixin modules: base, navigation, welcome, dependencies, network, environment, database, cluster, services, security, ollama, supabase, docker, build, packaging, deploy, validate, cfg. __init__.py composes ProleInstaller from all mixins and re-exports has_display(), main() for full backward compatibility.

All 37 tests pass with no regressions.
2026-02-20 14:43:20 -08:00