Legal

Service Level Statement

Most hosts publish a "99.9% SLA" with a credit process nobody ever successfully uses. We would rather tell you exactly what you are buying.

Effective 31 July 2026 Version 1.0 Part of the Terms of Service

1There is no uptime guarantee

Say it plainly

Blossom Host does not offer a contractual uptime guarantee and does not offer service credits for downtime. No plan includes one. If you need a contractual availability figure with financial remedies, we are not the right host and we will tell you so before you buy.

What we offer instead is a target we actually run the business to, and honest disclosure of where the weak points are. Everything on this page is what we believe to be true about the platform as operated today.

We target high availability on the datacenter and ISP lines and we treat unplanned downtime as an incident to be fixed immediately. That is an operating standard, not a warranty, and nothing here creates a contractual commitment or a right to a credit or refund.

2What we do commit to

These are real obligations, in the Terms:

  • Provision what you paid for. If we cannot deliver the service, you get a full refund.
  • Deliver the specification sold. The vCPU, memory, storage and IP allocation on your order is what your server gets.
  • Give notice before terminating without cause — 30 days, plus a pro-rata refund.
  • Follow the published expiry ladder before destroying anything, and warn you at each stage.
  • Tell you the truth about performance. We publish measured figures and retract numbers we cannot stand behind.

3Shared resources — read this before quoting our numbers

Except on dedicated bare metal, you are sharing physical hardware and network capacity.

  • An advertised 10 Gbps figure is the node's uplink, shared between every server on that node. It is not a dedicated or guaranteed per-server rate. Under contention your share is a fraction of it.
  • Uplinks may be asymmetric. Outbound capacity is often materially lower than inbound.
  • The ISP line is gigabit. Any single connection caps at approximately 1 Gbps regardless of aggregate site capacity.
  • Monthly transfer allowances are fair-use figures, oversubscribed against node capacity. They are enforced by a rate limiter and a monthly meter.
  • Latency, throughput, and benchmark figures we publish are measurements taken at a point in time. They are guidance, not commitments, and your results will differ.

We would rather sell you a larger plan or the Priority Bandwidth add-on than throttle you. If you are consistently hitting a ceiling, talk to us.

4Residential seats — beta, and explicitly so

Residential seats carry no availability or throughput commitment of any kind, and the disclosures in Terms section 4.2 are material to the price. In particular:

  • Customer traffic reaches the seat through a relay, and the relay — not the box — is the limiting factor. Throughput measured inside the box is not what you receive. We publish both figures and label which is which.
  • The relay is currently a single point of failure. We are working on that; until it is done, a relay outage is a fleet outage.
  • The underlying machines are rented on residential connections and can go away at short notice, taking the seat with them.
  • The underlying IP address changes. Your Blossom hostname is stable; do not depend on the IP.

We can withdraw a beta service at any time. If we do, we refund the unused portion of your prepaid term.

5Backups

No plan includes a backup you can rely on. Unless you have purchased the Backup & Recover add-on, assume no restorable copy of your server exists. Any operational disaster-recovery copies we hold are for our own continuity and are not a customer-facing service.

Where you have purchased Backup & Recover, we make reasonable efforts to maintain the stated restore points, but restoration is a reasonable-efforts obligation and not a guaranteed outcome. See Terms section 11.

Keep your own off-server copy of anything you cannot afford to lose. This is the single most common way customers lose data, and it is entirely avoidable.

6Maintenance

We perform planned maintenance — host patching, firmware, network changes, migrations — and schedule it for low-traffic periods where we can. We give advance notice through Discord, Telegram, and the dashboard when we expect customer-visible impact.

Emergency maintenance — a security patch, failing hardware, or an upstream requirement — happens without notice when waiting would be worse than acting. Planned and emergency maintenance are not treated as downtime for any purpose.

7Outside our control

We are not responsible for interruptions arising from:

  • anything you did inside your server — misconfiguration, a bad update, a full disk, a firewall rule locking you out, resource exhaustion, or software you installed;
  • your server being compromised, or our suspension of it under the Acceptable Use Policy;
  • suspension for non-payment or the expiry ladder running its course;
  • upstream provider, datacenter, transit, marketplace, or platform failure, action, or discontinuation;
  • denial-of-service attacks against you or against infrastructure you share, including null-routing an address to protect the node;
  • a third party's decision to block, throttle, blocklist, or refuse traffic from an address we assigned you; or
  • any force majeure event under Terms section 19.

8Support

Support runs through Discord, the Telegram bot, dashboard tickets, and email. There is no guaranteed response time and no 24/7 coverage. In practice most questions get answered quickly, because a person reads them rather than a ticket system.

We support the platform we operate — provisioning, networking, billing, hardware. We do not administer, secure, or debug what you run inside your server. We may help anyway, as a courtesy and without warranty.

9When something breaks

Our commitment is procedural rather than financial. In an incident affecting multiple customers we will: acknowledge it publicly in Discord rather than wait for tickets; say what is affected and what is not; give an honest estimate, including "we do not know yet"; and say afterwards what happened.

What we will not do is quietly absorb an outage and hope nobody noticed.

10Status and measurement

The latency board on our site shows live measured round-trip times from our production nodes. It is a measurement tool, not a status page, and it is not a commitment.

Incident communication happens in Discord. That is the fastest place to find out whether something is us or you.