drop old replica LB field and use useLoadBalancer for migration#3106
drop old replica LB field and use useLoadBalancer for migration#3106FxKu wants to merge 12 commits into
Conversation
FxKu
added
the
major
FxKu
requested review from
Jan-M,
idanovinda,
jopadi,
mikkeloscar and
sdudoladov
as code owners
| return nil | ||
| } | ||
|
|
||
| func (c *Cluster) syncMigrationService() error { |
There was a problem hiding this comment.
It would make sense to already pass a context here and use that in the relevant calls inside the function.
There was a problem hiding this comment.
I think in all our API calls we only use context.TODO() everywhere
|
As a general comment, the migration service is very specific to Zalando EKS migration and Zalando Clusters, it's not reasuable by anyone from opensource and it needs to stay until the migration is complete and possible longer if we make migration a more general feature. We have to understand and accept this overhead. The alternative solution would be to have the spilo-migrator create the service as part of the migration and bind it to the postgresql resource, that way the Zalando specific migration stuff would be located in one place outside of the postgres-operator. |
3 participants
time to finally remove the old load balancer flags after it was marked as deprecated for years. We will temporarily use the useLoadBalancer field to created an internal NLB which we use to migrate clusters between different K8s infras