5 Commits

Author SHA1 Message Date
51933afea6 feat(infra): Refactor Swarm networking for reliable DNS and stack ownership
Moves `iklimco-net` overlay network creation to be managed by the Docker Swarm stack, ensuring reliable embedded DNS resolution for inter-service communication. This resolves issues where services on external overlay networks failed to discover each other via Docker DNS.

This refactoring includes:
*   Removing the manual `iklimco-net` creation from the Ansible `swarm` role.
*   Adjusting `act_runner` configuration to connect job containers to `iklimco-net` only after the stack has deployed and created the network.
*   Setting `storagebox_file_mode` to `0600` for DB nodes to prevent "too open" errors with MongoDB keyfiles.
*   Provisioning dedicated bind mount directories for MongoDB and PostgreSQL on DB nodes with correct ownership and permissions.
*   Updating documentation to reflect the consolidated stack and network changes.
2026-05-26 01:08:12 +03:00
6f9d0d1588 feat(infra): Standardize StorageBox permissions and refactor DB stack name
- Ensure consistent directory and file permissions on StorageBox mounts for improved container access across application and database services.
- Introduce application-specific `storagebox_uid`/`gid` variables for more granular ownership control.
- Enhance StorageBox mount reliability by adding systemd reload and remount handlers for configuration changes.
- Add root credentials to Patroni's etcd configuration for authenticated communication.
- Update all relevant documentation and deployment scripts to use the `iklimco` Docker stack name for database services.
- Re-encrypt production vault secrets to include the new etcd password.
2026-05-23 18:11:01 +03:00
9e20f2fcf8 chore(prod): capture production bootstrap access configuration
Document and commit the production bootstrap state after the initial Hetzner and Ansible rollout.

- switch Ansible prod runbooks to use the shared vault password file

- record production admin CIDRs, SSH key path, encrypted group vault, and encrypted per-host vault files

- add generated production inventory and the prod setup history notes from the first bootstrap

- keep root password login disabled while preserving key-based root access for Ansible bootstrap continuity

- document separate Hetzner projects and tokens for test/prod and commit the prod provider lock file

- remove the private Redis firewall allowance from the prod Terraform firewall and matching setup docs
2026-05-19 17:49:59 +03:00
27f4f83f73 docs(prod): resolve cross-layer inconsistencies and complete prod env implementation
Ansible roles:
- act_runner/defaults: set act_runner_name to inventory_hostname (was
  hardcoded to iklim-test-app); added vault_gitea_runner_token to vault.yml
- prod/group_vars/all: restructured from flat files to all/ directory;
  added act_runner_labels override (prod-runner,ubuntu-24.04,hostname);
  added storagebox_managed_directories; added swarm_manager_ip and other
  prod-specific vars
- prod/roles/db_stack: prod-specific db_node tasks using StorageBox paths
  (/mnt/storagebox/db/...) instead of local paths
- docker/tasks: split firewalld loop into all-nodes (Swarm ports) and
  app-only (80/443) tasks
- swarm/tasks: added --advertise-addr private_ip to join commands for
  correct multi-homed node advertisement
- hardening/tasks: corrected firewalld drop zone configuration
- node_dirs/tasks: added /opt/iklimco/vault/data for Vault Raft volume
- db_stack/tasks/app_node: updated stale comment (removed pg-proxy reference)
- db_stack/templates: removed pg-proxy and mongo-proxy service blocks
- test/host_vars/iklim-app-01: added act_runner_name override to preserve
  existing test runner registration

Roadmap and setup docs:
- roadmap/03-infra-stack-changes: added replicas:0 for etcd/postgresql/
  mongodb/pg-proxy/mongo-proxy in prod overlay; updated placement table;
  fixed grafana/data mkdir (auto-created by Ansible); translated Turkish
  note to English
- roadmap/08-deploy-pipeline-update: updated stale "remains idle" note
  for standalone etcd (now disabled with replicas:0)
- roadmap/01-swarm-init-multinode: consistency fixes
- setup/06: added Outputs section and etcd firewall port documentation
- setup/07: removed prometheus/data from StorageBox acceptance criteria;
  replaced manual StorageBox mkdir section with Ansible auto-creation note;
  updated prod README section with full bootstrap instructions and vault docs;
  added act_runner_labels prod policy
- setup/08: extensive rewrite — aligned with Patroni etcd overlay DNS,
  corrected hcloud_firewall.app reference, updated all StorageBox paths
  from /prod/db/ to /db/
- setup/09: removed prometheus/data from acceptance criteria; updated
  runner label policy (removed docker/swarm-manager labels); added
  acceptance criterion for disabled services absent from docker service ls

Terraform:
- prod/firewall.tf: added missing DB subnet mutual rules (etcd, Patroni)
- prod/outputs.tf: added prod_floating_ip and prod_private_ips outputs
- prod/servers.tf: aligned placement group and naming
- prod/variables.tf: corrected variable descriptions
- prod/terraform.tfvars.example: updated defaults
- terraform/hetzner/README.md: new comprehensive README covering both
  test and prod environments with firewall tables and inventory instructions

ansible/README.md: expanded prod section with inventory groups, bootstrap
  run order, runner label policy, and vault variable documentation
2026-05-18 19:17:56 +03:00
f4b7f49968 chore: prepare prod ansible and db operations
Add the Ansible README and expand prod bootstrap coverage for StorageBox keys, DB labels, DB stack configuration, and act runner setup. Update MongoDB configuration for replica set support and refresh prod roadmap/setup documentation for Swarm labels, StorageBox-backed cert paths, and recovery guidance.
2026-05-15 20:39:57 +03:00