Andrew Pryde

e.g. [[1]](https://app.wercker.com/Oracle/mysql-operator/runs/end-to-end-tests/5b9fa7f6c81a8a00066d8a96?step=5b9fa81e82df1e00070eb386) [[2]](https://app.wercker.com/Oracle/mysql-operator/runs/end-to-end-tests/5b9fa62a8c38a200070e28c3?step=5b9fa640f07af00007e89b6d)

Backups should not be written to disk temporarily

3

Currently, the `mysqldump` backup provider writes the backup to the local filesystem and then opens the file and returns that as the `io.Reader` for the object storage client. https://github.com/oracle/mysql-operator/blob/2a251f7c0886a8302a63f59a34858047ea88127f/pkg/backup/executor/mysqldump/provider.go#L88-L108 We...

Instead of using the root user for GR we should create a group replication user. See: https://dev.mysql.com/doc/refman/8.0/en/mysql-innodb-cluster-production-deployment.html#mysql-innodb-cluster-user-privileges

Kubernetes 1.11 supports the [status subresource](https://kubernetes.io/docs/tasks/access-kubernetes-api/custom-resources/custom-resource-definitions/#status-subresource) on custom resources defined via CRDs. We should explore whether we can utilise this without breaking backwards compatibility.

Kubernetes 1.11 introduces support for the [scale subresource](https://kubernetes.io/docs/tasks/access-kubernetes-api/custom-resources/custom-resource-definitions/#scale-subresource). This allows us to support `kubectl scale` in the operator.

From Kubernetes v1.9 validation of custom objects is enabled by default via defining [OpenAPI v3](https://github.com/OAI/OpenAPI-Specification/blob/master/versions/3.0.0.md#schemaObject) schema. We should define schema for all our CRDs. See: https://kubernetes.io/docs/tasks/access-kubernetes-api/extend-api-custom-resource-definitions/#validation

## FEATURE REQUEST Implement (config map) leader election so that only one instance of the MySQL operator is running at once. Will need a little bit of thinking re support...

We should make use of [controller expectations](https://github.com/kubernetes/kubernetes/blob/20f27a77eeeb7c2ddc3e1e530c1c3d87cf05e494/pkg/controller/controller_utils.go#L107) in controllers where we create/delete resources. See [`ReplicaSetController`](https://github.com/kubernetes/kubernetes/blob/20f27a77eeeb7c2ddc3e1e530c1c3d87cf05e494/pkg/controller/replicaset/replica_set.go#L90) in core. The code hasn't yet made its way into client-go though (https://github.com/kubernetes/client-go/issues/332) so we...

Versioning

1

It would be preferable to have the role versioned (GH releases and on Ansible Galaxy) so that users can be confident that changes aren't going to affect their projects silently.

Incorrectly logged error: 'worker unable to determine worker memory usage'

2

4ac42695ce60cdf4786ca36b0cba170a3eb2e12f introduced a bug in `billiard.pool.Worker` when `max_memory_per_child` is set. The problem is that the [`else` block](https://github.com/celery/billiard/commit/4ac42695ce60cdf4786ca36b0cba170a3eb2e12f#diff-42cbdfeee4eeeb2688454e975c2d18ffR403) in the code below will always be executed when the worker's RSS is...