Deploying Trains Server: Kubernetes Using Helm

Trains is now ClearML

This documentation applies to the legacy Trains versions. For the latest documentation, see ClearML.

Prerequisites

A Kubernetes cluster.
kubectl is installed and configured (see Install and Set Up kubectl in the Kubernetes documentation).
helm is installed (see Installing Helm in the Helm documentation).
One node labeled app=trains.

Trains Server deployment uses node storage

If more than one node is labeled as app=trains, and you redeploy or update later, then Trains Server may not locate all your data.

Deploying

Securing deployment

By default, Trains Server deploys as an open network. To restrict Trains Server access, follow the instructions in the Network and security section, on the "Configuring Trains Server" page.

Step 1: Modify Elasticsearch default values in the Docker configuration file

Before deploying Trains Server in a Kubernetes cluster, you must modify several Elasticsearch settings in the Docker configuration. For more information, see Install Elasticsearch with Docker in the Elasticsearch documentation and Daemon configuration file in the Docker documentation.

To modify Elasticsearch default values in your Docker configuration file:

Connect to the node in the Kubernetes cluster that you labeled app=trains.
If your system contains a /etc/sysconfig/docker Docker configuration file, then edit it and add the options in quotes to the available arguments in the OPTIONS section:
```
OPTIONS="--default-ulimit nofile=1024:65536 --default-ulimit memlock=-1:-1"
```

If your system does not contain a /etc/sysconfig/docker Docker configuration file, then create or edit a /etc/docker/daemon.json file and add or modify the defaults-ulimits section as the following example shows:

{
    "default-ulimits": {
        "nofile": {
            "name": "nofile",
            "hard": 65536,
            "soft": 1024
        },
        "memlock":
        {
            "name": "memlock",
            "soft": -1,
            "hard": -1
        }
    }
}

Elasticsearch requires that the vm.max_map_count kernel setting, which is the maximum number of memory map areas a process can use, is set to at least 262144.

For CentOS 7, Ubuntu 16.04, Mint 18.3, Ubuntu 18.04 and Mint 19.x, we tested the following commands to set vm.max_map_count:
```
echo "vm.max_map_count=262144" > /tmp/99-trains.conf
sudo mv /tmp/99-trains.conf /etc/sysctl.d/99-trains.conf
sudo sysctl -w vm.max_map_count=262144
```
Restart docker:
```
sudo service docker restart
```

Step 2. Deploy Trains Server in the Kubernetes using Helm

After modifying several Elasticsearch settings in your Docker configuration (see Step 1), you can deploy Trains Server.

To deploy Trains Server in Kubernetes using Helm:

Add the trains-server repository to your Helm:

helm repo add allegroai https://allegroai.github.io/trains-server-helm/

Confirm the trains-server repository is now in Helm:
```
helm search trains
```
The helm search results must include allegroai/trains-server-chart.
Install trains-server-chart on your cluster:
```
helm install allegroai/trains-server-chart --namespace=trains --name trains-server
```
A trains namespace is created in your cluster and trains-server is deployed in it.

Port Mapping

After Trains Server is deployed, the services expose the following:

API server on 30008.
Web server on 30080.
File server on 30081.

The node ports map to the following container ports:

30080 maps to trains-webserver container on port 8080
30008 maps to trains-apiserver container on port 8008
30081 maps to trains-fileserver container on port 8081

We recommend using the container ports (8080, 8008, and 8081), or a load balancer (see the next section, Accessing Trains Server).

Accessing Trains Server

To access your Trains Server:

Create a load balancer and domain with records pointing to Trains Server using the following rules which Trains uses to translate domain names:
- The record to access the Trains Web-App (UI):
```
*app.<your domain name>.*
```
For example, trainsapp.mydomainname.com points to your node on port 30080.
- The record to access the Trains API:
```
*api.<your domain name>.*
```
For example, trainsapi.mydomainname.com points to your node on port 30008.
- The record to access the Trains file server:
```
*files.<your domain name>.*
```
  For example, trainsfiles.mydomainname.com points to your node on port 30081.

Upgrading

Use the current release

We strongly encourage you to keep your Trains Server up to date, by upgrading to the current release.

Upgrade using new or upgrade values.yaml

helm upgrade trains-server allegroai/trains-server-chart -f new-values.yaml

If you previously deployed a Trains Server, you must first delete old deployments using the following command:
```
helm delete --purge trains-server
```
If you are upgrading from Trains Server version 0.15 or older, a data migration is required before you upgrade. First follow these data migration instructions, and then continue this upgrade.

Upgrade your deployment to match repository version.

helm upgrade trains-server allegroai/trains-server-chart

Next Step

Configuring Trains for Trains Server.