Step 1: define the split of work and the number of instances
Start by recording that the client provides the machines, the network and the disks, and the software vendor deploys the Docker containers. Set up two instances: a production instance and a twin pre-production instance. Two instances mean two sets of resources, so the decision about a host or reserved resources should be taken at the start.
Enter in the protocol the person on the client's side responsible for the host, the network and the backups. The lack of this person is the most common cause of delays in the first week of the installation.
Step 2: prepare the host and the container environment
The full stack of the system consists of several services: a TLS reverse proxy, a PHP 8.4 application with Apache, a MySQL 8.0 database, an S3-compatible object store, Redis, a PDF generation service and a .NET helper service. In the development environment a mail catcher is added. The data sits in named volumes.
For the administrator this gives a list of items to prepare: a container runtime environment on the host, volumes for the database and the file store, and space for backups. The processor and memory resources should be sized to the number of users and events. Specific values are set after a conversation about the scale of the plant.
- Docker container environment on the host
- Volumes for the MySQL database and the S3-compatible object store
- Backup space on the client's disks
- Resources for the PDF service and the helper service
Step 3: lay out the network zones
The pattern for an on-premise installation covers four zones: the plant area with BLE transmitters, the users in the plant, the plant LAN and the client's private cloud in which the server runs. Before the installation, describe which subnets correspond to which zones.
The location works without GPS and without an external map service, so there is no need to open outbound access to such services. This limits the list of outbound rules and simplifies agreements with the security department.
Step 4: prepare the handover point
Traffic from outside, meaning phones on the GSM network and the vendor's service team, arrives only through the handover point under the client's control: a web application firewall (WAF), one published address, an outbound firewall and the end of the IPsec tunnel on the client's side.
Before the installation prepare a certificate for the published address, a WAF rule and the address of the tunnel endpoint. Record who may switch off the service tunnel. That decision lies with the client.
Step 5: check the supporting services and the proof of readiness
The first run of a clean environment, from an empty database to a working login, took about 8 minutes in the test, mainly because of the 380 migrations. Plan this in the installation schedule, so that the time is not read as a failure.
The proof of readiness is a short record: image versions, container status, the result of the login, the backup location and the date of the trial restore. The record goes into the acceptance protocol and is the basis for the audit at the client's site.
Video
Frequently asked questions
How many instances need to be prepared?
Two: a production instance and a twin pre-production instance. Both run in Docker containers on the client's infrastructure.
Do outbound connections to map services have to be opened?
No. The location works without GPS and without an external map service.
How long does the first start from an empty database take?
About 8 minutes in the test, mainly because of the 380 migrations.
Who decides about the access of the service team?
The client, because the end of the IPsec tunnel sits on the client's side.
Let's talk about the installation in your server room
We will present the requirements and the split of responsibilities for your plant based on the installation documentation. Book a conversation with the InnerWeb team.
Let's talk