* sync all resources to cluster fields (CronJob, Streams, Patroni resources)
* separated sync and delete logic for Patroni resources
* align delete streams and secrets logic with other resources
* rename gatherApplicationIds to getDistinctApplicationIds
* improve slot check before syncing streams CRD
* add ownerReferences and annotations diff to Patroni objects
* add extra sync code for config service so it does not get too ugly
* some bugfixes when comparing annotations and return err on found
* sync Patroni resources on update event and extended unit tests
* add config service/endpoint owner references check to e2e tes
* feat(498): Add ownerReferences to managed entities
* empty owner reference for cross namespace secret and more tests
* update ownerReferences of existing resources
* removing ownerReference requires Update API call
* CR ownerReference on PVC blocks pvc retention policy of statefulset
* make ownerreferences optional and disabled by default
* update unit test to check len ownerReferences
* update codegen
* add owner references e2e test
* update unit test
* add block_owner_deletion field to test owner reference
* fix typos and update docs once more
* reflect code feedback
---------
Co-authored-by: Max Begenau <max@begenau.com>
* Annotate PVC on Sync/Update, not only change PVC template
* Don't rotate pods when only annotations changed
* Annotate Logical Backup's and Pooler's pods
* Annotate PDB, Endpoints created by the Operator, Secrets, Logical Backup jobs
Inherited annotations are only added/updated, not removed
* add unit test and documentation for finalizers
* error msg with lower case and cover sync case
* try to avoid adding json-patch dependency
* use Update to remove finalizer
* changing status and finalizer during create
* do not call Delete() twice
* Add Finalizer functions to Cluster; add/remove finalizer on Create/Delete events
* Check if clusters have a deletion timestamp and we missed that event. Run Delete() and remove finalizer when done.
* Fix nil handling when using Service from map; Remove Service, Endpoint entries from their maps - just like with Secrets
* Add handling of ResourceNotFound to all delete functions (Service, Endpoint, LogicalBackup CronJob, PDB and Secret) - this is not a real error when deleting things
* Emit events when there are issues deleting resources to the user is informed
* Depend the removal of the Finalizer on all resources being deleted successfully first. Otherwise the next sync run should let us try again
* Add config option to enable finalizers
* Removed dangling whitespace at EOL
* config.EnableFinalizers is a bool pointer
---------
Co-authored-by: Felix Kunde <felix-kunde@gmx.de>
* Add support for pooler load balancer
Signed-off-by: Sergey Shatunov <me@prok.pw>
* Rename to enable_master_pooler_load_balancer
Signed-off-by: Sergey Shatunov <me@prok.pw>
* target port should be intval
* enhance pooler e2e test
* add new options to crds.go
Co-authored-by: Sergey Shatunov <me@prok.pw>
* init error arrays correctly
* avoid nilPointer when syncing connectionPooler
* getInfrastructureRoles should return error
* fix unit tests and return type for getInfrastructureRoles
* Create cross namespace secrets
* add test cases
* fixes
* Fixes
- include namespace in secret name only when namespace is provided
- use username.namespace as key to pgUsers only when namespace is
provided
- avoid conflict in the role creation in db by checking namespace
alongwith the username
* Update unit tests
* Fix test case
* Fixes
- update regular expression for usernames
- add test to allow check for valid usernames
- create pg roles with namespace (if any) appended in rolename
* add more test cases for valid usernames
* update docs
* fixes as per review comments
* update e2e
* fixes
* Add toggle to allow namespaced secrets
* update docs
* comment update
* Update e2e/tests/test_e2e.py
* few minor fixes
* fix unit tests
* fix e2e
* fix e2e attempt 2
* fix e2e
Co-authored-by: Rafia Sabih <rafia.sabih@zalando.de>
Co-authored-by: Felix Kunde <felix-kunde@gmx.de>
* replace statefulset on annotation diff
* remove update annotation function for statefulset
* add unit test for syncing annotations
* add inherited annotation to unit test
* add TODOs for moving rooling update label on pods
* steer rolling update via pod annotation
* rename patch method and fix reading flag on sync
* pass only pods to recreatePods function
* do not take address of iterator if you use it later
* add e2e test and pass switchover targets to recreatePods
* add wait_for_pod_failover for e2e test
* add one more e2e test case
* helm chart remove 1.6.0 archive from 1.6.0 archive
* reflect code review feedback
* Enable connection pooler for replica
* Refactor code for connection pooler
- Move all the relevant code to a separate file
- Move all the related tests to a separate file
- Avoid using cluster where not required
- Simplify the logic in sync and other methods
- Cleanup of duplicated or unused code
* Fix labels for the replica pods
* Update deleteConnectionPooler to include role
* Adding test cases and other changes
- Fix unit test and delete secret when required only
- Make sure we use empty fresh cluster for every test case.
* enhance e2e test
* Disable pooler in complete manifest as this is source for e2e too an creates unnecessary pooler setups.
Co-authored-by: Rafia Sabih <rafia.sabih@zalando.de>
Co-authored-by: Jan Mussler <janm81@gmail.com>
* Improving end 2 end tests, especially speed of execution and error, by implementing proper eventual asserts and timeouts.
* Add documentation for running individual tests
* Fixed String encoding in Patorni state check and error case
* Printing config as multi log line entity, makes it readable and grepable on startup
* Cosmetic changes to logs. Removed quotes from diff. Move all object diffs to text diff. Enabled padding for log level.
* Mount script with tools for easy logaccess and watching objects.
* Set proper update strategy for Postgres operator deployment.
* Move long running test to end. Move pooler test to new functions.
* Remove quote from valid K8s identifiers.
* allow using both infrastructure_roles_options
* new default values for user and role definition
* use robot_zmon as parent role
* add operator log to debug
* right name for old secret
* only extract if rolesDefs is empty
* set password1 in old infrastructure role
* fix new infra rile secret
* choose different role key for new secret
* set memberof everywhere
* reenable all tests
* reflect feedback
* remove condition for rolesDefs
* delete secrets the right way
* make a one function
* continue deleting secrets even if one delete fails
Co-authored-by: Felix Kunde <felix.kunde@zalando.de>
deleteConnectionPooler function incorrectly checks that the delete api response is ResourceNotFound. Looks like the only consequence is a confusing log message, but obviously it's wrong. Remove negation, since having ResourceNotFound as error is the good case.
Co-authored-by: Christian Rohmann <christian.rohmann@inovex.de>
* Initial commit
* Corrections
- set the type of the new configuration parameter to be array of
strings
- propagate the annotations to statefulset at sync
* Enable regular expression matching
* Improvements
-handle rollingUpdate flag
-modularize code
-rename config parameter name
* fix merge error
* Pass annotations to connection pooler deployment
* update code-gen
* Add documentation and update manifests
* add e2e test and introduce option in configmap
* fix service annotations test
* Add unit test
* fix e2e tests
* better key lookup of annotations tests
* add debug message for annotation tests
* Fix typos
* minor fix for looping
* Handle update path and renaming
- handle the update path to update sts and connection pooler deployment.
This way no need to wait for sync
- rename the parameter to downscaler_annotations
- handle other review comments
* another try to fix python loops
* Avoid unneccessary update events
* Update manifests
* some final polishing
* fix cluster_test after polishing
Co-authored-by: Rafia Sabih <rafia.sabih@zalando.de>
Co-authored-by: Felix Kunde <felix-kunde@gmx.de>
* Protected and system users can't be a connection pool user
It's not supported, neither it's a best practice. Also fix potential null
pointer access. For protected users it makes sense by intent of protecting this
users (e.g. from being overriden or used as something else than supposed). For
system users the reason is the same as for superuser, it's about replicastion
user and it's under patroni control.
This is implemented on both levels, operator config and postgresql manifest.
For the latter we just use default name in this case, assuming that operator
config is always correct. For the former, since it's a serious
misconfiguration, operator panics.
Connection pooler support
Add support for a connection pooler. The idea is to make it generic enough to
be able to switch between different implementations (e.g. pgbouncer or
odyssey). Operator needs to create a deployment with pooler and a service for
it to access.
For connection pool to work properly, a database needs to be prepared by
operator, namely a separate user have to be created with an access to an
installed lookup function (to fetch credential for other users).
This setups is supposed to be used only by robot/application users. Usually a
connection pool implementation is more CPU bounded, so it makes sense to create
several pods for connection pool with more emphasize on cpu resources. At the
moment there are no special affinity or tolerations assigned to bring those
pods closer to the database. For availability purposes minimal number of
connection pool pods is 2, ideally they have to be distributed between
different nodes/AZ, but it's not enforced in the operator itself. Available
configuration supposed to be ergonomic and in the normal case require minimum
changes to a manifest to enable connection pool. To have more control over the
configuration and functionality on the pool side one can customize the
corresponding docker image.
Co-authored-by: Felix Kunde <felix-kunde@gmx.de>
Previously, the operator put pg_hba into the bootstrap/pg_hba key of
Patroni. That had 2 adverse effects:
- pg_hba.conf was shadowed by Spilo default section in the local
postgresql configuration
- when updating pg_hba in the cluster manifest, the updated lines were
not propagated to DCS, since the key was defined in the boostrap
section of Patroni.
Include some minor refactoring, moving methods to unexported when
possible and commenting out usage of md5, so that gosec won't complain.
Per https://github.com/zalando-incubator/postgres-operator/issues/330
Review by @zerg-junior
Run more linters in the gometalinter, i.e. deadcode, megacheck,
nakedret, dup.
More consistent code formatting, remove two dead functions, eliminate
naked a bunch of naked returns, refactor a few functions to avoid code
duplication.
* Allow configuring pod priority globally and per cluster.
Allow to specify pod priority class for all pods managed by the operator,
as well as for those belonging to individual clusters.
Controlled by the pod_priority_class_name operator configuration
parameter and the podPriorityClassName manifest option.
See https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/#priorityclass
for the explanation on how to define priority classes since Kubernetes 1.8.
Some import order changes are due to go fmt.
Removal of OrphanDependents deprecated field.
Code review by @zerg-junior
There are shortcuts in this code, i.e. we created the deepcopy function
by using the deepcopy package instead of the generated code, that will
be addressed once migrated to client-go v8. Also, some objects,
particularly statefulsets, are still taken from v1beta, this will also
be addressed in further commits once the changes are stabilized.
Call Patroni API /config in order to set special options that are
ignored when set in the configuration file, such as max_connections.
Per https://github.com/zalando-incubator/postgres-operator/issues/297
* Some minor refacoring:
Rename Cluster ManualFailover to Swithover
Rename Patroni Failover to Switchover
Add more details to error messages and comments introduced in this PR.
Review by @zerg-junior
* Depreate old LB options, fix endpoint sync.
- deprecate useLoadBalancer, replicaLoadBalancer from the manifest
and enable_load_balancer from the operator configuration. The old
operator configuration options become no-op with this commit. For
the old manifest options, `useLoadBalancer` and `replicaLoadBalancer`
are still consulted, but only in the absense of the new ones
(enableMasterLoadBalancer and enableReplicaLoadBalancer).
- Make sure the endpoint being created during the sync receives proper
addresses subset. This is more critical for the replicas, as for the
masters Patroni will normally re-create the endpoint before the
operator.
- Avoid creating the replica endpoint, since it will be created automatically
by the corresponding service.
- Update the README and unit tests.
Code review by @mgomezch and @zerg-junior
Felix Kunde
Polina Bungina
Christian Rohmann
Philipp B
Rafia Sabih
Jan Mussler
Dmitry Dolgov
Eric
Sergey Dudoladov
Maxim Ivanov
Oleksii Kliukin
Oleksii Kliukin
Oleksii Kliukin