Has anyone actually asked RackNerd about that 847MB figure? I ran something similar on a Contabo box last year and it turned out they were reserving 128MB for a memory balloon that wasn't visible to the guest. Not defending it, just saying there might be a boring explanation before we get to "malicious liars"
Their control panel reporting 100% available while you're actively being oom-killed is pretty funny though. That's just bad telemetry.
Member
Eight days and nobody wants to touch the actual number
I keep a spreadsheet of this stuff from provider TOS pages. Most of them have stopped even pretending. "Fair share" is the new "dedicated resources" and it means even less.
847MB on a 1GB guarantee with 100% reported is just funny at this point. Not ha-ha funny.
Member
Sixteen days late but yeah, that dashboard metric is doing a lot of heavy lifting for them. "100% available" until you actually try to touch it.
We had a similar bug years back where the panel pulled from the host node's free memory instead of the container's cgroup limit. Looked great, meant nothing. Took a customer running a proper memory burn to catch it.
Dunno if RackNerd's is the same root cause or just optimistic UI design. Either way, agree with you — that's not density, that's denial.
Member
This is the bit that gets me. You can bury anything in terms of service and claim transparency. Hostinger at least puts it up front in the plan name which is more than most.
23 days late but I wanted to add: the 847MB thing from RackNerd is probably ballooning kernel overhead or a cgroup limit set below physical allocation. Not defending it, just saying the 100% panel reading is likely "of your quota" not "of the box." Still misleading.
Has anyone actually seen a provider publish live density numbers. Not marketing fluff, like a real dashboard
Been 24 days and I'm still thinking about this
That's the thing though, even when it's there it's buried in 4pt grey-on-grey legal text. "Fair share" means nothing until you're the one getting throttled at noon on a Tuesday
I'd actually respect RackNerd more if their panel said "84% memory available (847MB/1024MB, node at 91% capacity)". Same broken reality, different relationship with the customer
The three quid argument only works because nobody knows what the real cost baseline is anymore
Yeah I think the control panel just pulls from /proc/meminfo without checking cgroups limits or whatever theyre using to slice the node, so it reports the host total
Still running that little malloc bomb on other providers when I cant sleep, results are patchy
Contabo and Vultr both killed me around 920MB on "1GB" plans which feels more honest than RackNerds 847MB plus fantasy dashboard
52 days later and I still havent cancelled that RackNerd box, its become a weird hobby now
Member
56 days but this one stuck with me
That's the real trick though isn't it? The disclosure buried in paragraph seven of a TOS that hasn't been updated since 2014. Hostinger at least front-loads it, most shops just hide behind "fair share" and call it a day
I still want to know what RackNerd's control panel is actually measuring when it says 100%. Free memory? Committed? Some internal quota that doesn't map to reality? The gap between the number and the experience is where the actual dishonesty lives
Member
79 days but this one's still relevant
That's the bit that gets me, the dashboard lying in real time
We publish our density ranges on the Hetzner site, nobody clicks it but it's there. "Budget VPS: 60-90 instances per compute pool" etc
The 100% available thing is worse than silence though. At least silence lets you assume the numbers are aspirational
Has anyone actually seen a provider update density stats live? Like a traffic light or something
Member
89 days, still thinking about this thread. Lee_mcr has the right of it, I reckon. The disclosure matters more than the ratio.
I used to work support at a place that did the honest oversell dance. We had the numbers in a tooltip on the order page. Maybe 1 in 200 tickets ever referenced it, and those were usually people who *liked* knowing the density. The rest just wanted cheap compute that worked.
Still. 847MB from a "1GB guaranteed" plan and a dashboard that says 100%? That's not a disclosure problem, that's a telemetry problem. Or a "we don't know what our own hypervisor is doing" problem.
Either way I'm not sure which stings more.
112 days, but this still annoys me.
Did you ever check if that was kernel overhead or if RackNerd was actually billing for the full gig? I've seen both. Some providers count cache/buffers as "used by system" and don't charge you, others just pocket the difference.
I ran the same malloc test on a KnownHost box last year and got exactly 1024MB. Their dashboard showed 89% used. At least it was honest.
