Availability, measurement and service credits
This document describes four things: what availability we commit to per product class, how we measure it, which periods are not counted and what happens if we do not meet the commitment. It specifies section 10 of the Terms and Conditions and is part of every contract for the listed products.
Version 4.2, valid from 1 July 2026 · previous version 4.1 from 1 October 2025 · reference: section 10 Terms and Conditions.We deliberately avoided making the exclusions so broad that nothing is left in the end: emergency maintenance counts as half an outage for us, and the burden of proof for an exclusion reason lies with us.As a fileService Level Agreement Version 4.2, valid from 1 July 2026 · PDFAll documents for procurement
- Highest commitment
- 99.99%
- 4 min 19 s per month
- Lowest commitment
- 99.90%
- 43 min 12 s per month
- Maximum service credit
- 100%
- of the monthly base fee
- Application deadline
- 30 days
- Exclusion period
Three classes, twelve products, one calculation rule
The committed availability depends on the class, not the product name. The permitted downtime is based on a calendar month of 30 days (43,200 minutes); in months with 28, 29 or 31 days, it shifts accordingly.
| Availability class | Committed value | Downtime per month | Products | Measurement point |
|---|---|---|---|---|
| Class A | 99.99% | 4 min 19 s | Backbone and transit, Anycast DNS, highly available Kubernetes Control Plane | Edge router of the region, authoritative response per zone, API server of the cluster |
| Class B | 99.95% | 21 min 36 s | Cloud instances, Object Storage, Block Storage, highly available managed databases, Load Balancer | Network connection and control plane (the management layer through which resources are created and queried), S3 endpoint, block device, connection endpoint |
| Class C | 99.90% | 43 min 12 s | Dedicated servers, VPS, standard Kubernetes Control Plane, web hosting | Uplink interface of the server, network connection of the instance, HTTP response |
Availability classes, committed value and permitted downtime per calendar month
Class A
- Committed value
- 99.99%
- Downtime per month
- 4 min 19 s
- Products
- Backbone and transit, Anycast DNS, highly available Kubernetes Control Plane
- Measurement point
- Edge router of the region, authoritative response per zone, API server of the cluster
Class B
- Committed value
- 99.95%
- Downtime per month
- 21 min 36 s
- Products
- Cloud instances, Object Storage, Block Storage, highly available managed databases, Load Balancer
- Measurement point
- Network connection and control plane (the management layer through which resources are created and queried), S3 endpoint, block device, connection endpoint
Class C
- Committed value
- 99.90%
- Downtime per month
- 43 min 12 s
- Products
- Dedicated servers, VPS, standard Kubernetes Control Plane, web hosting
- Measurement point
- Uplink interface of the server, network connection of the instance, HTTP response
For Object Storage, an additional durability commitment of 99.999999999% per object and year applies. Durability and availability are different metrics: an object can exist and still be temporarily inaccessible.
Per resource, per month
Availability is calculated separately for each individual resource and each calendar month. There is no offsetting across multiple resources, multiple months or multiple locations — neither in your favour nor in ours.
Change of class
Class C applies to Kubernetes clusters without a highly available control plane. You make the selection when creating the cluster; a subsequent change is possible and takes effect from the following calendar month.
Order of precedence in case of conflict
Individual written agreement, then this Service Level Agreement (SLA), then the Terms and Conditions, then the service description. Products in beta are excluded from any commitment.
Class A applies to backbone and transit — 99.99% per month, i.e. 4 minutes 19 seconds of downtime. Measurement takes place at the edge router of the region, i.e. the router where our network hands over to third-party networks. What happens beyond that in the third-party network does not count against this commitment, but can still affect you.
Service Level Agreement · Version 4.2, valid from 1 July 2026
Measured from the outside, not by us
Measurements are taken every 60 seconds by 14 checkpoints in at least five independent networks. No checkpoint is located in our own network or at a location where the platform runs. A measurement interval is only considered disrupted if at least three checkpoints receive no valid response in the same interval — this ensures that disruptions on individual measurement paths are not attributed to you or us.
The availability for a month is calculated as total minutes minus disrupted minutes, divided by total minutes, rounded to three decimal places. We retain the raw data for 24 months; if a service credit request is rejected, we provide it unprompted.
- Measurement interval
- 60s
- Checkpoints
- 14in 5 networks
- Threshold per interval
- 3Checkpoints
- Retention
- 24months
The measurements of the last 90 days per service are publicly accessible. We also publish values there that trigger a service credit. To the status page.
What counts as a valid response
For each product, it is precisely defined which response a checkpoint must see. Everything else counts as a disruption.
- Bare Metal and VPS: ICMP echo and a complete TCP connection setup at the uplink interface.
- Cloud instances: additionally a successful status query via the control plane.
- Object Storage: Writing, reading and deleting a 4 KiB test object. Responses with HTTP 5xx or a latency over 5 seconds count as errors.
- Block Storage: successful read and write access to the device with a latency below 250 ms.
- DNS: authoritative response to a test query per zone with correct signature.
- Load Balancer: HTTP response with status code below 500 at the virtual IP address.
In the event of a hardware defect on a dedicated server, the attributable downtime begins with the report in the ticket system or with our own detection by our monitoring — whichever is earlier. It ends with the restoration of accessibility, not with the recovery of your data.
What is expressly not covered by the commitment
An availability commitment is only as good as its list of exceptions. That is why this list is provided here in full and not in a footnote. The burden of proof for the existence of an exclusion lies with us: if we claim an exclusion, we will name it in the individual case and justify it.
Announced maintenance
Provided it was announced at least five calendar days in advance and does not exceed eight hours per calendar month and resource. Any time beyond this counts fully as downtime.
Force majeure
In particular natural events, war, riot, official orders, strikes by third parties and a failure of the public power grid lasting more than 72 hours.
Customer fault
Incorrect configuration, exceeding agreed quotas, errors in your own software or software used by you, accidental deletion of your own resources and a justified suspension according to § 7 of the Terms and Conditions.
Disruptions outside our network
That is, beyond our edge routers, including disruptions in your access network or with third-party transit providers on the way to you.
Attack mitigation
If the volume exceeds our filtering capacity and a null routing measure becomes necessary — the attacked IP address is temporarily removed from the network so the rest can continue running. Limited to four hours per event. You will be notified of the measure immediately.
Beta products
Products marked as beta in the customer panel and price list. No availability is guaranteed for them.
Payment default
After an unsuccessful reminder with a notice period of at least ten calendar days.
Operating system and application
For dedicated servers, the reachability of the network connection is measured, not that of the operating system. A server that no longer responds due to a faulty configuration is not considered to have failed.
Emergency maintenance counts as half
Maintenance to avert an acute threat to operations, data or security can take place without notice. In deviation from the list above, it is calculated as 50% downtime, and we will state the reasons in writing within 48 hours.
We consider it dishonest to completely exempt ourselves behind the term “emergency”. Anyone who declares every unplanned job as emergency maintenance ends up with an empty commitment.
The tier system, and what it measures
If the promised availability is not met in a calendar month, you receive a service credit towards the recurring monthly base fee of the affected individual resource. The downtimes in the table are calculated from the achieved percentage; they are not an additional condition.
| Service credit tier | Achieved value in the calendar month | Downtime per month | Service credit | Example at €189.00 net base fee |
|---|---|---|---|---|
| Level 1 | below the commitment, at least 99.500% | up to 3 h 36 min | 10% | €18.90 net |
| Level 2 | below 99.500%, at least 99.000% | 3 h 36 min to 7 h 12 min | 25% | €47.25 net |
| Level 3 | below 99.000%, at least 98.000% | 7 h 12 min to 14 h 24 min | 50% | €94.50 net |
| Level 4 | below 98.000% | more than 14 h 24 min | 100% | €189.00 net |
Service credit tiers — Share of the monthly net base fee of the affected resource, example at €189.00
Level 1
- Achieved value in the calendar month
- below the commitment, at least 99.500%
- Downtime per month
- up to 3 h 36 min
- Service credit
- 10%
- Example at €189.00 net base fee
- €18.90 net
Level 2
- Achieved value in the calendar month
- below 99.500%, at least 99.000%
- Downtime per month
- 3 h 36 min to 7 h 12 min
- Service credit
- 25%
- Example at €189.00 net base fee
- €47.25 net
Level 3
- Achieved value in the calendar month
- below 99.000%, at least 98.000%
- Downtime per month
- 7 h 12 min to 14 h 24 min
- Service credit
- 50%
- Example at €189.00 net base fee
- €94.50 net
Level 4
- Achieved value in the calendar month
- below 98.000%
- Downtime per month
- more than 14 h 24 min
- Service credit
- 100%
- Example at €189.00 net base fee
- €189.00 net
The tier system applies to each resource individually. If multiple resources are affected, the service credits are added up; the total per resource is limited to 100 percent of its monthly fee. The downtimes refer to a calendar month of 30 days.
Assessment basis and limits
- Included
- Exclusively the recurring monthly base fee of the affected resource.
- Not included
- Setup fees, consumption-based charges, charges for additional services of other resources, and taxes.
- Final
- The service credit is the final compensation for the shortfall. Statutory claims arising from intent or gross negligence, from injury to life, body and health, or under the Product Liability Act remain unaffected.
- Payout
- Service credits appear on the next invoice. A payout is only made at the end of the contract with no outstanding invoice amount, then within 30 days.
Claims procedure and cut-off period
If our own measurement shows a shortfall, we issue the service credit without an application. It appears on the invoice of the following month with the resource, measured value, and applied tier level. If you notice a shortfall that our measurement did not capture, request the service credit via a ticket in the “SLA service credit” category.
Required information
- Identifier of the affected resource and region
- Start and end of the disruption with date, time and time zone
- Description of the error, with log excerpts where possible
- Your own measurement data, if available — it is helpful, but not a prerequisite
- Application deadline
- 30 calendar days
- Review by us
- 10 working days
- Second review
- 14 days
The application deadline of 30 calendar days after the end of the affected month is a preclusion period: After its expiry, there is no longer any entitlement to a service credit. In the event of rejection, we provide the measurement data for the period without you having to request it. If you do not agree with the assessment, a person not involved in the initial decision will review it again.
Initial response yes, resolution time no
The initial response time is the time between the receipt of your report and the first substantive reply from a responsible person. An automated acknowledgement of receipt is not an initial response. We do not commit to a resolution time. How long the fix takes depends on the cause — and any commitment to this would be a number we could not keep.
| Support level | P1 critical | P2 significant | P3 minor | P4 request | Availability |
|---|---|---|---|---|---|
| Basic (included in the price) | 8 h | 12 h | 24 h | 48 h | Mon–Fri 08:00–18:00 |
| Business | 2 h | 4 h | 8 h | 24 h | P1 24/7, otherwise Mon–Sat |
| Enterprise | 30 min | 2 h | 4 h | 12 h | 24/7, telephone |
| Critical | 15 min | 1 h | 2 h | 8 h | 24/7, dedicated phone number |
Initial response times per support level and priority
Basic (included in the price)
- P1 critical
- 8 h
- P2 significant
- 12 h
- P3 minor
- 24 h
- P4 request
- 48 h
- Availability
- Mon–Fri 08:00–18:00
Business
- P1 critical
- 2 h
- P2 significant
- 4 h
- P3 minor
- 8 h
- P4 request
- 24 h
- Availability
- P1 24/7, otherwise Mon–Sat
Enterprise
- P1 critical
- 30 min
- P2 significant
- 2 h
- P3 minor
- 4 h
- P4 request
- 12 h
- Availability
- 24/7, telephone
Critical
- P1 critical
- 15 min
- P2 significant
- 1 h
- P3 minor
- 2 h
- P4 request
- 8 h
- Availability
- 24/7, dedicated phone number
For the Enterprise and Critical levels, we designate fixed contacts; for Critical, an additional technical contact who knows your environment.
- P1 — critical
- Production operations are completely down, there is no workaround.
- P2 — significant
- Production operations are significantly impaired or a workaround can only be used with noticeable effort.
- P3 — minor
- A single function is disrupted, operations essentially continue.
- P4 — request
- Questions about usage, change requests, quotes, billing.
You determine the classification when reporting. We only correct it after consultation and with justification; a silent downgrade does not take place.
Fixed windows, hard limit
A maintenance window is a previously announced period during which we work on a facility. Planned work takes place exclusively within these windows and is limited to eight hours per month and resource. If a system has only one power supply or network connection, maintenance on the power or network infrastructure can interrupt it — even if the facility itself is designed redundantly. We inform affected accounts individually in advance.
- Fixed windows
- Tue and Thu, 01:00–05:00We carry out plannable work exclusively in these windows, in the local time of the affected location.
- Notice without interruption
- 5 calendar daysBy email to the registered technical contacts and simultaneously on the status page.
- Notice with interruption
- 14 calendar daysIncludes scope, expected impact and the affected resources.
- Upper limit per month
- 8 hoursPer calendar month and resource. Anything beyond this counts fully as downtime.
- Postponement on request
- up to 14 daysFor dedicated resources, request at least 48 hours before the window. Security-relevant work is excluded.
- Emergency maintenance
- counts at 50%Permitted without notice, but counts as half downtime. We will explain the reasons in writing within 48 hours.
We will announce changes to this document to your detriment in text form at least 90 calendar days before they take effect; if you object within 30 days, the previous SLA will continue to apply to existing contracts until the end of the current contract period. Changes exclusively in your favour take effect without a notice period. We provide previous versions on request; each version bears a version number and validity date.
Related documents
ENTRONYX CLOUD is a demonstration project. This text is simulated and has no legal effect; in particular, it does not replace a legal review.
The numbers are public — even the ones that cost us money
The status page shows the measured availability of the last 90 days for each service from the same measurement that decides on service credits. If you need a different commitment, you negotiate it individually: we do not write a number into a contract that we cannot keep.
- Measurement interval
- 60 s
- Checkpoints
- 14 in 5 networks
- Raw data
- 24 months
- Emergency maintenance
- counts at 50%