improve cost efficiency for clusters-config

[enekofb]

We are currently in about 3k month. this ticket to review and apply actions we could take to reduce costs.

[enekofb]

i guess that some very basic approach based on the factors in the picture:

Reduce kubernetes engine costs by:

• delete unused clusters
• merge used clusters

Reduce compute engine costs by:

• downsize used clusters
• schedule active hours for compute
• delete unused instances or orphan

Networking:

• identify paid networking flows
• move them to free of charge or
• delete the ones that we dont need

[enekofb]

delete unused clusters

resources only in eu-north-1

asked to clusters candidates to delete https://weaveworks.slack.com/archives/C045Z5FNHNE/p1696954046409449
but not many though

[enekofb]

merge used clusters

No clusters to merge

[enekofb]

delete unused instances or orphan

we could see some -> deleting the ones without a cluster associated

deleted

[enekofb]

Networking: delete the ones that we dont need
Seen elastics ips and natgateways ... sympthon of waste

after clean up

[enekofb]

@morancj have just done a bunch of deletions that we should see impact in the next days ... however i cannot access to billing

shout with the costs update when you see (tomorrow or after tomorrow)

[enekofb]

Actions to do around Reduce compute engine costs by:

• downsize used clusters -> reduce number of subnets to 2 for some availability
• schedule active hours for compute -> set default schedule time

[morancj]

@morancj have just done a bunch of deletions that we should see impact in the next days ... however i cannot access to billing

Nor me with AdministratorAccess, likely the SCP´s in engineering_account.tf.

[morancj]

likely the SCP´s in engineering_account.tf.

Access should be fixed by https://github.com/weaveworks/corp-infra/pull/67. PTAL @enekofb and LMK if you now have access to:

• AWS Billing Dashboard
• Cost Explorer

[morancj]

Billing & Cost access confirmed fixed via Slack

[enekofb]

Reduction of 2/3 -> no more actions just yet

Next:

• downsize used clusters -> 2 nodes
• schedule active hours