InnerWeb
Skip to main content

Isolated instance in Docker: host requirements on the plant side

An isolated instance with its own database and its own email is a setup in which the IT department is responsible for the host and the network, not for the application code. This guide collects what IT must provide on the plant server and points out the decisions worth recording before the first start.

21 / h
event limit per error type
Source: karta
89 / h
log entry limit per instance
Source: karta

How the instance is built

Each client gets an isolated instance with its own database and its own email system. The instance runs in Docker containers, including on the client's own server, and some environments run on Kubernetes. Isolation means that the plant's data does not share a database with other clients, and that a failure of one instance does not spread to the others.

The configuration is described by an environment variable contract without values: a list of the names the instance accepts. Values, including secrets, are supplied by the runtime environment. For IT this is the only formal point of contact with the application configuration.

Host requirements

The requirements are ordinary for a container environment: a server with a Docker engine installed (or a Kubernetes cluster), space for images, the database and backups, a clock synchronized with a time server, and a service account for starting containers. Specific resource parameters are set individually, because they depend on the number of users and devices.

The host clock matters beyond tidiness: audit timestamps use the Europe/Warsaw time zone from October 2026, and a time gap between the host and external sources makes event analysis harder.

  • Docker engine or Kubernetes cluster in a version supported by the vendor.
  • Time synchronization with the plant's NTP server.
  • Space for images, the database and backups.
  • A service account with limited permissions for the containers.
  • A DNS name and a certificate for the instance address.

Error reporting limits and transport time

Error reporting limits and transport timeBar chart: limit of 21 events per hour per error type and 89 per hour per instance020406080100events per hourPer error type: 21 events per hour21Per error typeLog entries per instance: 89 events per hour89Log entries per instance
Source: Component card: error reporting documentation

Network and outbound traffic

When the client's network blocks all outbound traffic from the application host, alarm webhooks are stored permanently and an external poller collects them through an inbound channel. This matters for IT, because outgoing connections do not have to be opened just so that an alarm arrives.

It is worth deciding before go-live whether the host's outbound traffic is limited to indicated addresses or fully closed. In the second case the collection of alarms has to be planned and the path from the event to the notification has to be tested.

Secrets, email and backups

Instance secrets are kept outside the repository, in the plant's mechanism or in environment variables, with a rotation plan. Email is separate for each instance, and the shared sending address sits in the InnerWeb domain, with SPF, DKIM and DMARC records unchanged. IT only checks how the plant's email treats messages from that domain.

Database backups should go outside the instance host and have a recorded restore time. The audit trail covers 107 entities changed by a person, and the data tables of machine origin are excluded from it, so the size of the backup and of the log can be predicted.

Monitoring and host acceptance

Error reports go to one shared Sentry project, with a limit of 21 events per hour per error type and 89 per hour per instance for log entries. The transport has a limit of 1 second per connection and 2 seconds per response, so an unavailable receiver does not slow the application down. The release version is the short commit hash of the deployed image.

At host acceptance IT records the image version, the result of the alarm test, the result of the email test and the location of the backups. Such a set is the audit evidence that the environment was started deliberately.

Video

Frequently asked questions

Can the instance run on Kubernetes?

Yes, some environments run on Kubernetes, while the basic model is Docker containers, including on the client's own server.

Does the host need internet access?

Not always. With blocked outbound traffic, alarm webhooks are stored permanently and a poller collects them through an inbound channel.

Where are the environment variable values stored?

Outside the repository, in the plant's runtime environment mechanism. The variable contract describes only their names and meaning.

Sources

Let's check your host before installation

We will go through server requirements, network and backups with your IT team and decide whether Docker is enough or a cluster is the better choice.

Let's talk