* build multi-arch pooler image
* add pooler build step in delivery.yaml and bump pooler version
* pull from docker hub not zalando registry
* add pooler step to ghcr workflow
* pass infra roles to auth file via pooler entrypoint
* introduce extra pooler secret for mounting auth_file
* use pbgouncer as image name and push to ghcr on next merge
* build with latest pgbouncer
* integrate new image in e2e process and update pooler image default
* update pooler build dependencies
* build pooler image for e2e test
* more Makefile and e2e run script tweaking
---------
Co-authored-by: Ida Novindasari <idanovinda@gmail.com>
* add fix to recreate non running pods in syncStatefulsets
* remove TestSyncStatefulSetNonRunningPodsDoNotBlockRecreatio
* revert pod_test
* pod without status
---------
Co-authored-by: Felix Kunde <felix-kunde@gmx.de>
Co-authored-by: Ida Novindasari <idanovinda@gmail.com>
read_stored_clusters and read_versions build their S3 clients with
endpoint_url=AWS_ENDPOINT, but read_basebackups used a bare
client('s3') for both the list_objects_v2 paginator and the per-key
get_object call. On MinIO / S3-compatible backends the list+get
requests go to the default AWS endpoint, so the Backups tab renders
cluster/version prefixes (picked up by the correctly-configured
read_stored_clusters) but then returns empty base backup details
(silently no hits against the real backend) (#3078).
Build s3_client once per call with endpoint_url=AWS_ENDPOINT and reuse
it for both the paginator and get_object. No behaviour change when
AWS_ENDPOINT is unset; boto3 defaults to the AWS endpoint either way.
Fixes#3078
Signed-off-by: SAY-5 <SAY-5@users.noreply.github.com>
Co-authored-by: SAY-5 <SAY-5@users.noreply.github.com>
Co-authored-by: Ida Novindasari <idanovinda@gmail.com>
* fix(logical-backup): wait for PG connectivity before running backup
The backup script connects to the target PostgreSQL pod immediately
after resolving its IP via the Kubernetes API. When NetworkPolicy is
enforced via iptables, a newly-created pod's IP may not yet be present
in the destination node's ingress allow lists, causing cross-node
connections to be rejected until the next policy sync.
This adds a pg_isready retry loop before the dump starts, with
configurable retries and delay via LOGICAL_BACKUP_CONNECT_RETRIES
(default: 10) and LOGICAL_BACKUP_CONNECT_RETRY_DELAY (default: 2s).
Signed-off-by: Zadkiel AHARONIAN <zaharonian@ccl-consulting.fr>
* docs: document LOGICAL_BACKUP_CONNECT_RETRIES and RETRY_DELAY env vars
Document the new environment variables that control the pg_isready
retry loop added in the previous commit. These are passed via the
existing logical_backup_cronjob_environment_secret mechanism.
Signed-off-by: Zadkiel AHARONIAN <zaharonian@ccl-consulting.fr>
---------
Signed-off-by: Zadkiel AHARONIAN <zaharonian@ccl-consulting.fr>
Co-authored-by: Ida Novindasari <idanovinda@gmail.com>
* Updates Needed for WASM Target
* switch to regular (instead of local) build flags
* update codegen to match other scripts
---------
Co-authored-by: Felix Kunde <felix-kunde@gmx.de>
* Add support for pg18 and remove pg13
* Update general spilo image and use new rebuilt e2e spilo image
---------
Co-authored-by: Polina Bungina <polina.bungina@zalando.de>
* bugfix: fix image naming to use it in e2e test
* Fix embeded crd
* Add crd generation as dependency for docker
---------
Co-authored-by: Polina Bungina <polina.bungina@zalando.de>
* Sort postgresql.crd.yaml
* Generate postgresql CRD from go structs
* Expand sidecars, env and initcontainers
* Embed CRD to be submitted by the operator
Signed-off-by: Mikkel Oscar Lyderik Larsen <mikkel.larsen@zalando.de>
---------
Signed-off-by: Mikkel Oscar Lyderik Larsen <mikkel.larsen@zalando.de>
- Add ignore_slots field to Patroni struct in CRD
- Add ignore_slots to patroniDCS struct for Spilo configuration
- Generate ignore_slots in SPILO_CONFIGURATION when specified
- Update CRD manifest to accept ignore_slots field
- Add unit test for ignore_slots configuration
This enables PostgreSQL 17 native slot synchronization support by allowing
users to configure Patroni to ignore specific replication slot types (e.g.,
logical slots) during failover operations.
Users can now configure ignore_slots in their PostgreSQL manifest:
patroni:
ignore_slots:
- type: logical
This instructs Patroni to ignore logical replication slots during failover,
which is essential for PostgreSQL 17's native logical slot synchronization
feature where slots are automatically synced to standbys.
* fix docker build for UI and bumped some outdated versions in docs and config
* update helm chart image again because of wrong format field
* switch to new registry ghcr.io for e2e test
* update e2e test runner Dockerfile
* fix switchover schedule tests
Previously the tests would fail depending on the local time zone and the
time of day the test was being run.
---------
Co-authored-by: Felix Kunde <felix-kunde@gmx.de>
Co-authored-by: Mikkel Oscar Lyderik Larsen <mikkeloscar@users.noreply.github.com>
* separate pipeline for internal/external build and support multi arch
* remove distutils and use node:lts-alpine in both
* change base image arguments
* fix local naming and base image
* address feedback
---------
Co-authored-by: Felix Kunde <felix-kunde@gmx.de>
Felix Kunde
annielzy
Polina Bungina
Andreas Mårtensson
Sai Asish Y
Zadkiel AHARONIAN
dependabot[bot]
DDD
Ali Algohary
Ida Novindasari
Jorge Solorzano
Zavel Paytsev
Mikkel Oscar Lyderik Larsen
Pavel Zaytsev
Steven Berler