Noah Potash

I've nearly made this work, with the following changes: **Partitionspec field names need to match the columns** On line 154 of pkg\operator\aws_usage_hive.go Change the from this ``` p := metering.HiveTablePartition{...

> If I built a custom reporting-operator image would you be able to try it out to see if it resolves the issue for you? Yeah I'm happy to help...

I fired up the fix-aws-datasources1 reporting-operator image. As expected (since I made the same changes) it behaves just like mine does, getting far enough to fail on creating the partition...

I used the hive cli (beeline) to manually run ``` ALTER TABLE datasource_metering_aws_billing ADD IF NOT EXISTS PARTITION (`billing_period_start`='20191001', `billing_period_end`='20191101') LOCATION 's3a://my-redacted-billing-bucket/cur-report/cur-report/20191001-20191101/d0abd31d-5d9b-4af1-80e9-8596bcc7b6e3/'; ``` and it worked, so I'm not sure...

If it's a clue, while the table _did_ exist in hive, it appeared to be empty (but did have a ton of columns). Did you push the fsGroup change to...

I used the env vars, this is included in the output of ./hack/upstream-install.sh: ``` using $METERING_OPERATOR_IMAGE_REPO=quay.io/ecnahc515/origin-metering-ansible-operator to override metering-operator image using $METERING_OPERATOR_IMAGE_TAG=fix-aws-datasources1 to override metering-operator image tag ``` Here's my...

Seems like we're super close. I ran into one easily fixed issue with the reporting operator: `could not list retrieve AWS billing report keys: WebIdentityErr: failed to retrieve credentials\ncaused by:...

Thanks for the quick response. > 1. Are you seeing something other than this? If not, is the deprovisioning process too quick? I haven't actually tried it yet. I'm only...

There are a few different implementations discussed in aws/karpenter-core#753. If you're referring to https://github.com/aws/karpenter-core/issues/753, preventing _all_ nodes from deprovisioning until a certain time, irrespective of drift detection (if I understood...

If I'm understanding correctly, I could create a DisruptionBudget configured to disallow deprovisioning from drift (I assume drift would be among the "reasons you wouldn't want to see your nodes...