Servers & Cloud
Know what is about to go wrong on your servers, and fix it safely.
InfraNest watches every server across your clouds and tells you what needs you. It shows each fix before it runs. It also finds what you pay for and never use.
Free plan · No credit card · Connect a provider in 2 minutes
- One list of what needs you, across every cloud
- Reboots that keep your site up, with the plan shown first
- A warning days before a disk fills up
- A clear list of what you can stop paying for
Needs you · 7
api-01
A port is reachable without need
Compares the firewall rules with what the agent reads listening on the server and who connects to it. A rule open for nothing, a database listening on every address, or a port only private addresses use is attack surface you can close. The Networking tab lists every port.
TCP 80 · nothing listens behind the rule
Remove the rule6 more
Works with your clouds
Reboot a live server without your visitors noticing
A reboot is the job people put off, because nobody knows what it will take down. InfraNest shows the plan before you press the button. The server leaves its load balancer, and you see which server carries the visitors and how busy it will be. It reboots, and comes back only once the agent and your monitors say it is fine. If that does not happen, it stays out and your whole team is told. Pick the quietest hour and go to bed.
- Refuses to start if the site would go down
- A scheduled database reboot needs your yes first
- Load balancers are on Business and up
Reboot web-01 safely
An update has been waiting for this reboot since October 21, as the agent reports.
What happens
- 1Take web-01 out of web-lbVisitors go to web-02 meanwhile: healthy, and at about 27.2% CPU with its share.
- 2Reboot through HetznerA clean shutdown signal first; a hard reset only if it does not respond within 2 minutes.
- 3Wait for the agent to report a new bootA new boot id, and no reboot waiting any more.
- 4Wait for its monitors to be upEvery monitor that watches it, checked again since the reboot, within 10 minutes.
- 5Put it back into web-lbIf step 3 or 4 does not happen within 10 minutes, it stays out and everybody in the workspace is told.
The quietest time is Saturday 22:00, from 4 weeks of readings.
Reboot staging-web-01 safely
staging-web-01 is the only server behind staging-lb. Taking it out would take the site down, so it cannot start now. Add a second server behind staging-lb first, or reboot it in your provider’s panel or on the machine at a time the site can be down. If staging-web-01 itself is down, InfraNest reboots it where it is instead.
Server monitoring that knows what is normal for each machine
A number alone tells you little. 70% CPU during a nightly backup is fine, and the same 70% on a quiet afternoon is not. After four weeks, each server has its own normal range for every hour of the week. InfraNest then says in plain words what stands out, when the disk will be full and when the server is quietest. The agent makes this possible: one command, about a minute. Most clouds see a server only from the outside, so they cannot report memory or disk space. DigitalOcean can, if its own monitoring is on, and InfraNest shows those figures.
- CPU and network work with nothing installed
- New servers borrow the range of a similar server
- Unusual readings never replace your alert rules
db-01
Hetzner · ccx23
What stands out in this period
- · CPU was above its normal range for this time of the week for about 3 h.
- · Load rises and falls with network traffic: the server is most likely busy serving requests, not running away on its own.
CPU against its normal range
79%/var/lib/postgresql
db-01 runs out of disk on /var/lib/postgresql in about 21 days.
Quietest time: Saturday 22:00
Usually around 28% CPU then, against around 36% CPU on average.
From 4 weeks of this server’s own readings, so the day of the week counts.
Reboot safely thenPlan maintenance
Nothing here sends an alert. Your alert rules work as before.
Recommendations you did not have to ask for, each with the next step
Waste never breaks, so nothing ever warns you about it. A disk attached to nothing, an address left behind, a staging box nobody uses: each is billed every month. InfraNest finds them and shows the evidence and the cost, and nothing goes until you confirm it. It reads 30 days of peaks before it suggests a smaller size, so a server that averages 8% and hits 90% every night is left alone. The Security Advisor does the same for gaps, and each finding has a button for the fix.
- Final snapshots last 14 days, on Pro and up
- An address still used in DNS is not released
- Smaller-size advice needs the agent on that server
9 things cost €102.56 a month and do nothing
Each one is checked against DNS, monitoring, the agent and your load balancers. Nothing is removed until you confirm it here.
And 5 more in the review
Servers, and volumes where the provider can, get a final snapshot before they go, kept for 14 days. Addresses are checked against DNS first.
staging-web-01Could be one size smaller
s-2vcpu-2gb would save €6.00 a month.
Highest in 30 days: 11.2% CPU, 17.9% memory.
On s-2vcpu-2gb, the same peaks would be about 11.2% CPU and 35.8% memory, under the 60% left as headroom.
Resize to s-2vcpu-2gb…Keep, it is neededAsk me again in 30 days
web-01: Uses little of its size, but cpx21 has a 80 GB disk, and a disk cannot shrink: this server’s is 160 GB.
No firewall attached
Fix in firewallA firewall lets you expose only the ports you actually use. Without one, every port on this server is reachable from the internet.
No backups found
Open backupsAutomated backups let you recover from data loss, ransomware or a bad deploy. Enable them for anything you care about.
No snapshot taken
Create snapshotA snapshot is a quick, manual restore point — handy to take right before risky changes like an OS upgrade.
Find out your server’s address is blocked before your mail stops arriving
When a server’s address lands on a mail blocklist, your mail gets refused or ends up in spam. Most people learn this from an unhappy customer. InfraNest checks each server’s public IPv4 address against SpamCop and PSBL, every day on paid plans and every week on Free. You get an alert when it is listed and another when it clears. The Protection card puts this on the server’s page, next to its firewall, backups, delete protection and monitoring, so you can see what is covered at a glance.
- A new server is checked straight away
- Check again once a day after a removal request
- Start an automation when an address is listed
Protection
mail-01 · 198.51.100.166
- Firewall
- Needs the agent
- Backups
- 1 snapshot
- Delete protection
- Turn on
- Monitoring
- From the provider
- Alerts
- 1 rule
- Updates and reboots
- Needs the agent
- Mail blocklists
- On SpamCop
mail-01
Address is on a mail blocklist
Mail sent from 198.51.100.166 is refused or marked as spam by mail servers that use SpamCop. Find what sent spam from it, stop it, then ask the list to remove the address.
SpamCopCheck again
Your alert destinations hear when it is listed, and again when it clears.
A way back when a change goes wrong, or the server locks you out
Two questions decide how bad an afternoon gets: can I undo this, and can I still get in? Provider backups run on a schedule and restore in a click. Snapshots are the ones you take yourself, just before the risky thing. If the server stops letting you in, there is a console in your browser. It works like a screen and keyboard attached to the machine, and every session is in the audit log.
- Backups and snapshots are on Pro and up
- Snapshots show what they really cost
- Console logins expire shortly after use
Manual point-in-time copies you take yourself.
web-01 golden imageProtectedfrom web-01 · 80 GB disk · 6.2 GB stored6.2 GBdb-01 nightlyfrom db-01 · 40 GB disk · 11.1 GB stored11.1 GBshop-old before shutdownfrom shop-old · 80 GB disk · 21.5 GB stored21.5 GB9 snapshots, 3 protected from deletion — 655 GB of disk stored as 105.1 GB, 84% smaller.
For when SSH refuses, the server will not finish booting, or a firewall rule shut you out.
Credentials are temporary, requested per session and expiring shortly after. Opening one is written to the audit log, so the team can see who connected and when.
Everything else it handles
The parts that are only interesting when you need them.
No server gets deleted by accident
Delete destroys the machine and stops the bill. Stop tracking only takes it out of InfraNest. The confirmation says which one you are doing. Delete protection has two locks, ours and your provider’s where it has one, and either lock is enough to refuse.
What a server really costs
The plan price is only part of the bill. Each server adds up its backups, volumes, IP addresses and snapshots, plus anything you add yourself, like a licence. Prices are net of VAT and labelled an estimate. Cost history and the Spend report are on Business and up.
Alerts that wait for a real problem
“CPU above 85% for 15 minutes” means the average over those minutes, so a one-minute backup spike wakes nobody. A rule on memory or disk needs the agent, and stays greyed out until it is there. Each alert sends two messages at most: it fired, and it is back.
A month of capacity, side by side
Every server in one table with its peak next to its average. A steady 60% and one bad afternoon can have the same average, and only one is a reason to resize. A server with no readings says “No readings”, never zero.
Private networks
Every server on a private network can reach every other. Before you attach one, InfraNest names what becomes reachable and says whether your firewall can filter that traffic. Private networks are on Business and up.
Create a server without leaving InfraNest
Pick the account, then the name, size, location and image. The options come live from your provider, so you only see what you can order. A firewall can be attached as the server is created.
Only what your cloud can do
A button appears only if your provider supports it. Where a cloud cannot do something, the control is greyed out with the reason. You never press a button that can only fail.
Each server says what it is
Web server, database, cache, queue worker and more, recognised from what runs on it. The evidence is one click away, and your correction sticks. Automations that reboot servers leave a database alone unless you allow it.
Volumes
Create, attach, resize and delete across providers. A resize grows the disk, not the filesystem, so InfraNest gives you the exact command to run afterwards. On Pro and up.
Load balancers
See what each one listens on and which servers sit behind it. You can add or remove a target by hand or from an automation. On Business and up.
Reserved IPs
Every reserved address across your accounts, with what it points at, its reverse DNS and its monthly cost. Addresses attached to nothing are counted and totalled, because they still cost money.
Servers we do not manage
Bare metal or a box under a desk can be added and watched with the agent, marked Self-managed. It shows up under Monitoring, because nothing on the Servers page can reach it without a provider.
Automatable, and audited
Power, reboot, backups, delete protection, volumes and load balancer targets are all automation actions. Every change and every console session is in the audit log.
See what needs you in about two minutes
Connect one provider with an API token. Nothing changes until you say so.
Provider consoles vs InfraNest
The difference between four tabs and one answer.
Provider consoles
- Four consoles, and the full picture only in your head
- CPU graphs, and no warning until a disk is full
- Reboots at night, hoping the site stays up
- A bill that hides unused disks and addresses
- Blocked mail, found when a customer complains
With InfraNest
- One list of what needs you, across every cloud
- A normal range per server, and a date for a full disk
- A plan shown first, and a reboot that keeps visitors served
- A list of what to stop paying for, with the evidence
- An alert when an address lands on a mail blocklist
One login, ten modules
Nine modules are on every plan, Free included, and firewalls start on Pro. What changes is how much of each you get, plus a few advanced features from Pro.
Build on it
The cost of doing it separately
Buy each piece from a different tool and it adds up fast:
- Uptime monitoring
- ~€26
- SSL tracking
- ~€15
- Status page
- ~€26
- Across 3 separate tools
- ~€67/mo
Also in Pro, not counted above: 100 domains with DNS, 10 servers across your cloud providers, 25 automations, Dynamic IP and Drop Catch.
Each line is that tool’s cheapest paid plan. Pro covers all of it: 50 monitors, 100 certificates and 3 status pages on your own domain.
Prices checked October 2026 · dollar prices at the ECB rate
Built for how you work
Try it with a free tool
Frequently asked
Which clouds are supported?
Hetzner Cloud, DigitalOcean, TransIP and OVHcloud. What you can do differs by provider. Where a cloud cannot do something, the control is greyed out with the reason. OVHcloud VPS servers are ordered and cancelled in OVHcloud’s own panel, so InfraNest sends you there. The integrations page lists what works where.
Do I have to install anything to monitor my servers?
No. Connect a provider and you get CPU and network graphs, cost, backups and the full lifecycle. OVHcloud has no metrics API, so there the agent is the only source. The agent is optional: one command, about a minute. It adds memory, disk space, load, what is filling the disk, the state of your services and an alert when a server goes quiet.
Why can’t most providers report memory or disk space?
Those numbers live inside the machine, and a cloud sees it from the outside. It can measure CPU and network, but not how full your disk is. DigitalOcean is the exception: with its own monitoring on, it reports memory and disk space, and InfraNest shows them. Load, processes, services and alert rules on memory or disk still need the agent.
How do I reboot a server without downtime?
Put at least two servers behind a load balancer. InfraNest takes one out, reboots it, waits until the agent and your monitors say it is back, then puts it back in. It refuses to start if the server is the only one behind the load balancer, or if the others are unhealthy or too busy. Without a load balancer, the reboot still waits for the machine and its monitors, but what it serves is down meanwhile. Load balancers are on Business and up.
How do I know if my server’s IP is on a mail blocklist?
InfraNest checks each server’s public IPv4 address against SpamCop and PSBL, every day on paid plans and every week on Free. A new server is checked straight away, because providers reuse addresses. If it is listed, you get an alert, and the server’s page tells you what to do next. You get a second alert when it clears.
How does InfraNest find what I can stop paying for?
It looks for six things: idle servers, disks attached to nothing, unassigned addresses, old snapshots, empty load balancers and servers one size too big. Each is checked against DNS, your monitors, the agent and your load balancers. Nothing is removed until you confirm it.
Will alerts wake me for a one-minute spike?
No. A rule compares the average over its whole window, so a short spike does not reach the threshold. The average also has to come back about 10% under the threshold before anything is called recovered. You get two messages at most per alert.
What is the difference between deleting a server and stopping tracking?
Delete destroys the machine at your provider and stops the bill. Stop tracking leaves it running and billing, and only takes it out of InfraNest. It comes back with its history, tags and settings if you track it again.
See what needs you, and what you can stop paying for
Connect a provider and get your server list, your first recommendations and the savings review within minutes.
Free plan · No credit card required · Set up in minutes