Contabo "managed" VPS down, support claims no auto alerts fired

I feel this in my bones. Contabo "managed" means you manage while they cash the invoice. Since 2 years on their yearly plan, server down every weekend, no refund policy they say since march but downtime since forever.

Important to read fine print

They sell you managed but when MariaDB chokes at 3 AM it is your problem. Your 4.7G node with 2G per script? That is not upgrade, that is demolition. And their monitoring sleeping through whole business day? I seen this movie. My box was swapping to death, support ticket open 14 hours before human answer.

You got no root, so you cannot fix, but they blame "client responsibility" for PHP values their own cPanel push? Make it make sense. They overwrite your stable config then charge you for privilege of debugging their mess.

Retention promise 3 months, Advocacy deny it. Classic Contabo. Left hand sell you dream, right hand slap you awake.

I been telling forum since 2022: cheap yearly plan is expensive when weekend downtime cost you client. They not managed, they absent.

Since 2017 running CentOS 6 on InterServer box. PHP 5.4. Client saying "server working why change". I Important to them but they not Important to update. Finally hacked last month, they call me "site show casino ads". Now they listen. 8 years. 8 years.

Some host only learn when disaster knock. Contabo not even learn then.

I Have Learned This The Hard Vay With Hetzner. The Alerts They Promise Are The Fiction. I Vatch My Ovn Uptime No\v Because The Managed Means Nothing.

Backups on a €4/mo Contabo plan? Bro that's the fiction. You get what you pay for, and "managed" at these tiers means jack. I run my own rsync to a second box because I already bought three.

This hits close. We had a OVHcloud box a while back, "managed" with monitoring that never once pinged us when nginx ate itself. Found out from a client email, same as you.

The OOM loop for a full day though... that's special. I've seen MariaDB spiral on a small VPS, sure, but managed should mean someone gets paged. Not you paging them after the fact.

What gets me is the config handoff. If their automation pushed values, when does it stop being their problem and start being yours? We don't resell managed stuff precisely because that line's never clear in the contract. You think you're buying hands-on, you're buying hands-off until you beg.

Contabo's side would be useful, agreed, but I've been on the receiving end of those tickets. "Works on my end, must be your workload." Then you dig and find they doubled innodb_buffer_pool via unattended-upgrades three days prior.

Do they publish what "monitored" actually tracks? CPU average? Or individual service state? Big difference, and most of these shops won't show you the dashboard.

Managed at Contabo price points is basically a checkbox on a sales page. They'll patch nginx once a quarter and call it a day.

Idk why I do this but I stopped trusting anyone else's monitoring after my last "fully managed" box went dark for six hours. Now I just run a second $3 NAT instance with uptime-kuma pinging the first.

Actually the memory limit behavior is well-documented in cPanel's own implementation notes. Actually a 2G per-process ceiling on a 4.7G node isn't merely unwise, it's a configuration that violates basic multi-tenant resource isolation principles. Actually RFC 7838 doesn't help here but the cPanel documentation from 2022 explicitly warns against memory_limit exceeding 25% of available system RAM in managed environments.

Actually the "client responsibility" deflection is the standard Hetzner playbook (https://www.hetzner.com). I migrated a colleague's project away from their ecosystem in 2021 after identical finger-pointing during a botched EasyApache rollout. Actually their Level 3 designation appears to be primarily a title rather than a technical depth indicator.

Actually the swap exhaustion pattern you describe matches what I'd expect when OOM killer behavior is misconfigured alongside overcommitted memory. Actually MariaDB's memory allocator tends to be disproportionately targeted in such scenarios because of its mmap patterns. Actually if their monitoring stack relies on cPanel's built-in alerts rather than independent node-level telemetry, this blindspot is architectural, not incidental.

Actually 16 years of billing history should entitle you to escalation paths that bypass Dominic M.'s discretionary ceiling. Actually their retention specialists historically had broader compensation authority than frontline team leads. Actually I would reference the prior Dinda R. Commitment explicitly in any further correspondence.

Actually the no-root-access constraint makes their "client responsibility" framing particularly indefensible. Actually you cannot modify php.ini at the system level without WHM privileges, which they withhold. Actually this is a contractual contradiction, not a technical dispute.

I already said my piece but reading this again got my blood up. Two years of weekend crashes, no refund, and they still finding new ways to ghost you.

The part that cooks me is "managed" with no alerts. My Contabo box would go dark Saturday night, come back Monday like nothing happened. I open ticket, they say "we see no issues in logs." Logs! The server was unreachable for 36 hours, what logs you reading?

And this Dominic M. Using your cancellation date to close the case? Same energy. They got a playbook for making problems disappear instead of fixing them. Important to read fine print but also important to read reviews from people who already bled for you.

RackNerd treating you better? Good. I finally walked last month, went to a small shop where the owner answers WhatsApp at 2am. Costs more than yearly plan, sure. But my uptime went from "weekend lottery" to actual 99%.

Hetzner being same group explains lot. Same template, same script breaking your setup, same support script of "not our fault." These cheap yearly plans are traps. They price it so you eat the pain instead of migrating.

8 years I learned this lesson slow. You learned it fast. That is difference between us.

The way they closed your ticket using the cancellation date is the classic move. I have seen this with Hetzner too. The support finds any excuse to make the problem go away instead of fixing it.

I am not surprised the alerts did nothing. The 24/7 monitoring is the marketing text. What they actually monitor is the invoice payment. The server can be on fire and they will not notice until you open the ticket yourself.

The part about the vendor script is what gets me. You pay for managed so you do not have to worry about random blueprints being forced on your account. But they do it anyway and then pretend it is normal. It is not normal.

I am glad you moved to RackNerd. At least there you know where you stand. Self managed is honest. You do your own work or you pay extra. Contabo sells you the dream and delivers the headache.

The MariaDB under OOM should have been the easiest alert to catch. Any basic script would see the process dying. But they probably have the thresholds set so high that nothing triggers until the whole node is dead. Then they blame the client for using too much RAM.

I have my spreadsheet for uptime now. I do not trust any of them to watch my boxes. The managed label is just the checkbox like someone said. It means nothing when you actually need help.

You set it up yourself on Contabo "managed"? Buddy that's like bringing your own engine to a mechanic shop.

Yeah I get that angle, but the thread title literally says "managed" in quotes. We all know what shop we're talking about.

Still, your point about newbies grabbing self-managed stuff blind is real. Seen too many RackNerd customers open tickets asking why their "unmanaged" box doesn't patch itself. The marketing blurbs don't help — https://www.racknerd.com sure loves its "fully managed" banners on every page.

What monitoring did you end up rolling yourself, out of curiosity? Custom scripts or something off the shelf...