NEWJNB-01 Johannesburg is taking pre-orders — first 50 instances at launch pricing.
All systems operational
ParagonEST 2019
Sign inConfigure
Knowledge base

Written down once, so nobody types it twice.

Everything here started as a ticket. When the same answer went out for the third time we moved it into an article, ran the commands against a fresh instance, and linked it from the reply. That is the whole editorial process.

01

What is in here

Operational documentation for people who already have root and would like to be finished before lunch. There is no chapter explaining what a server is.

Debian 13 and AlmaLinux 10 unless an article says otherwise. Older releases usually work; the paths move.

Where a command differs between Debian and AlmaLinux, both are given rather than one with a vague note attached. Where something cannot be done at all, the article says so and stops. Nothing here is aspirational.

What you will not find is a tutorial on your own application. We run infrastructure. Configuring your reverse proxy is your business, right up to the point where the packets stop arriving — at which point it becomes ours, and there is an article about that too.

If a command in here fails on a current image, that is a bug in the article and we want the ticket. Quote the slug.

02

How it is filed

Six categories, chosen because they match the six kinds of ticket we actually receive rather than any tidy taxonomy of hosting.

01

Getting started

The first hour of an instance: login, cloud-init, API tokens. Read these once and most of the rest becomes optional.

02

Networking

Addressing, reverse DNS, the routed IPv6 /64, fair-use bandwidth, packet loss, and what happens while you are being attacked.

03

Storage

Extra NVMe volumes, filesystems, snapshots and restores. Short section, because block devices are not complicated when nobody is overselling them.

04

Security

Keys, nftables, two-factor authentication. Everything here is something we would do on our own machines.

05

Billing

Account balance, underpaid invoices, and the timeline between a missed renewal and a destroyed instance.

06

Operations

Custom media, the out-of-band console, plan changes, site moves, kernel tuning, and steal time that reads zero.

03

When an article does not answer it

The knowledge base covers what we can write down in advance. Anything specific to your instance, your traffic or your invoice needs a human, and there is one at every hour of the day.

  1. 01

    Check the status page

    If a site is having an incident, it is on /status before it is in your inbox. Start there when something worked yesterday and does not today.

  2. 02

    Try the guides

    Longer pieces that build something end to end, rather than answering one question. They live at /guides and assume you have read the relevant article here.

  3. 03

    Open a ticket

    Median first response is eleven minutes, measured across the last quarter at every hour. Tickets go to engineers, not to a queue that forwards them to engineers.

  4. 04

    Use the console

    If the instance is unreachable, the out-of-band console gets you a shell over a path that does not depend on your network being correct. It is included on every plan.

A ticket with an instance ID, a timestamp carrying a timezone and the exact command that failed is answered in one round trip. A ticket that says “my server is slow” is answered with a question.

Getting started

03
01First login, and the ten minutes after itWhat to do between receiving the root password and putting the instance on the public internet: check the host key, patch, make a user, harden sshd.Getting started
02Do the first-boot work with cloud-initAttach a user-data document to the order and the instance arrives with your user, your key, your packages and your firewall already in place.Getting started
03Create an API token and make the first callScoped tokens, the handful of calls you will use on day one, and the rate limits and error shapes worth knowing before you script against them.Getting started

Security

03
01Add an SSH key and switch passwords offGenerate an ed25519 key, install it properly, test it before you break anything, and turn password authentication off for good.Security
02A sensible nftables baselineA default-drop ruleset that permits what an instance genuinely needs, rate-limits SSH, and does not lock you out while you are still writing it.Security
03Two-factor authentication and recovery codesEnabling TOTP or a hardware key, why the recovery codes matter more here than anywhere else, and what happens if you lose all of them.Security

Networking

06
01Set reverse DNS on your addressesEditing PTR records for IPv4 and IPv6 from the panel, why mail servers care about the forward record, and how to confirm the change took.Networking
02Configure the routed IPv6 /64Putting your included /64 on the interface under Debian 13 and AlmaLinux 10, checking the route, and using the rest of the prefix for containers.Networking
03What the fair-use bandwidth figures meanTraffic is unmetered under a published sustained rate rather than a monthly cap; this is how it is measured and what happens if you cross it.Networking
04Find packet loss with MTR, and read it correctlyMost reported packet loss is an ICMP rate limit on a transit router; here is how to run MTR properly and tell the two apart before writing the ticket.Networking
05Run a Tor relay without breaking the acceptable use policyMiddle relays and bridges are welcome at any site and exits need a conversation first; configuration, bandwidth limits, and the abuse mail to expect.Networking
06What happens during a DDoS eventFiltering sits permanently in path, so most attacks end before anyone notices; what you see, what we do, and what we can tell you afterwards.Networking

Storage

02
01Attach an extra NVMe volume and mount itFinding the new block device, picking a filesystem, mounting it so a missing disk cannot stop the boot, and growing it later without downtime.Storage
02Take a snapshot, and put it backHow snapshots are taken, why crash-consistent is usually enough, how to quiesce when it is not, and what a restore does to a running instance.Storage

Operations

06
01Rebuild an instance from a custom ISOFetching your own installation media, booting it on the virtual console, and getting an unattended install to finish without anyone watching it.Operations
02Get in when the network is downThe serial and VNC console reaches the instance over a path that does not depend on your network configuration, which is precisely when you need it.Operations
03Steal time, and why yours reads zeroSteal time is the hypervisor admitting another guest took your core; on dedicated cores there is nothing to take, so the useful part is what to check instead.Operations
04Kernel settings for a lot of open connectionsThe sysctl values and systemd limits that matter when a box holds tens of thousands of sockets, and the ones copied from old blog posts that do nothing.Operations
05Move an instance to another siteDisk copies between cities are quick, but the addresses change; this is the sequence that keeps downtime to the copy and DNS to a minute.Operations
06Upgrade a plan with as little downtime as possibleCores, memory and disk can grow while the instance runs; a change of family cannot. What the guest has to do before it notices the new hardware.Operations

Billing

02
01Pay from an account balanceTopping up once and spending it down is quieter than paying an invoice on-chain every month, and it removes the renewal that waits on confirmations.Billing
02When an invoice is underpaid, overpaid or lateShort payments, late payments and payments to the wrong chain, in order of how often they happen and how easily each one is put right.Billing
Ready when you are

Ask a human instead

Support is staffed continuously and reads every ticket in the order it arrived. No tier one, no script, no invitation to reboot and try again.

Configure a serverTalk to an engineer