Kestrel Cloud

Their managed Postgres failover actually worked during our incident drill. Zero data loss, about 11 seconds of write pause.

Kubernetes upgrades happen without drama, and we get two weeks' notice with a rollback path every time.

Billing is the real feature. Every line maps to a metric we can query ourselves, and the invoice has matched the estimate three months running.

Billing is the real feature. Every line maps to a metric we can query ourselves, and the invoice has matched the estimate three months running.

Object storage latency is noticeably better than what we had before, and support answered our first ticket within an hour — on a Sunday.

Snapshot restores are fast enough that we test them monthly. A 4 TB volume came back in 22 minutes to the same region.