This. People think "alert" means "stop." It means "email you while your meter spins."
I got a $340 surprise from GCP because I set alert at 100% with no action. Took me 45 minutes to burn through free credits on a misconfigured load balancer. Alerts were cheerfully pinging my inbox the whole time.
Budget actions are the actual circuit breaker. Most providers hide this in submenus.
Worth adding: ask about reverse DNS delegation too. Some hosts let you keep the IP but move you to a different /24 with new SWIP. Your SPF still passes but receivers see "this IP never sent mail before."
OVHcloud kept my delegation exact same. Contabo made me re-request rDNS through ticket, took 4 days, mail sat in spam folders half that week.
If you're sending transactional mail, "same IP" is not enough. Need same prefix history, same ASN reputation, same delegation path. Ask specific questions, get specific answers in writing.
Tried Oracle. Twice. Both times account provisioning failed, had to file ticket, took 8 days first time, 13 second. Free tier is great if you can get in the door. YMMV heavily.
Their paid support is separate from account issues too. Stuck in free tier support queue forever.
Not saying don't try it. Saying "transparent upgrade path" includes "can actually create account without sacrificing goat." KnownHost at least took my money promptly.
Member
To bring this back to upgrade paths: their anycast is same network across plans. I started on VPS, upgraded to dedicated, same anycast announcement, same protection profile. No re-announcement lag, no migration window where you're exposed.
This is actually unusual. Most providers I've used, changing plan class means new network edge, new BGP session, 15-30 minute vulnerability window. OVHcloud's architecture treats it as capacity change, not topology change.
Not defending them generally. But for "grow without pain," this specific thing they do right. I used https://www.ovhcloud.com/en/vps/ when I started.
Has anyone used Hetzner for this? Their cloud VPS upgrades are in-place resize, same IPv4, no migration. But their dedicated root servers are separate product line — you can't "upgrade" from cloud to dedicated, it's new order, new everything.
So "transparent upgrade path" depends on what you're upgrading between. Within product line? Hetzner is smooth. Across product lines? Same forced migration as anyone.
I stayed on cloud VPS because of this. Outgrew the CPU but not enough to justify starting over.
All of you talking about "in-place resize" and "same IPv4" like these are features. They're table stakes. You pay for compute, you keep your addressing. Anything else is a billing event dressed as architecture.
Real transparent upgrade is hardware you own, colo contract you control. OVHcloud does colocation in same datacenters as their cloud. Move from rented to owned, same racks, same fiber. That's an upgrade path. Everything else is just changing your rental agreement.
Yes I'm expensive to be friends with.
Cool. My four servers cost less than your one month of colo probably. Different worlds.
Back to planet budget: I tried UpCloud, they do in-place resize with same IP, but only within same datacenter. I started in Singapore, wanted to move to Frankfurt for EU users — forced migration, new IP, exactly the problem Doug had.
"Transparent" has boundaries everywhere. Read the fine print on geography.
