Parameters
About this task
In the installation package, we include some sample YAML files (azure.yaml, google.yaml), which can be used as a starting point to configure the Kubernetes deployment.
For the complete list of parameters, please consult Values.yaml inside the chart sdl-mtedge-8.6.3.tgz.
Procedure
At a minimum, you need to modify the following parameters:
- mtedge.image.registry: URL prefix of your own registry.
- mtedge.ingress.host: The host name you plan to use for deploying for Language Weaver Edge.
Example:
# NFS volume provisioning. Only used with nfs-server-provisioner
persistence:
enabled: true # always set to true
storageClass: "standard" # the storage class the provisioner will use to allocate the initial pool
size: 100Gi # minimum 20 GB + 3GB per Language Pair
extraArgs:
device-based-fsids: "0"
# Edge deployment configuration
mtedge:
image:
registry: YOUR_REGISTRY # MANDATORY. The prefix of your Docker registry.
ingress:
host: YOUR_HOST # MANDATORY. the Ingress host
# the TLS configuration below is optional
tls:
enabled: true # TLS with automatic certificate renewal.
renew: # Certificate Manager must be installed (see above).
email: user@example.com # Must be a valid email address, as Let's Encrypt will validate it. The domain example.com will not work.
staging: false # With staging turned on, the certificate will be self-signed instead of being requested from Let's Encrypt.
storage:
controller:
size: 10Gi # storage used for Edge Controller 20GB is a safe value
class: mtedge-controller-storage-class # storage class used for ReadWriteOnce
lp:
size: 100Gi # shared storage used for Language Pairs
class: "nfs" # storage class. Must support ReadWriteMany
# memory and CPU limits for Controller, Translation Engine, Job Engine, Translation Engine pods.
resources:
controller:
requests:
cpu: 4
memory: 8G
limits:
cpu: 16
memory: 8G
tengine:
requests:
cpu: 4
memory: 3G
limits:
cpu: 4
memory: 3G
jengine:
requests:
cpu: 4
memory: 4G
limits:
cpu: 16
memory: 4G
trainingEngine:
requests:
cpu: 4
memory: 15G
limits:
cpu: 8
memory: 32G
# Job Engines
jengines:
replicas: 2 # the number of job engines. Must be at least one
# Translation Engines
tengines:
engchi: # any unique name for this Translation Engine
lp: engchi_generic_srv_tnmv_8_4_x_5 # LP name. Provided by Language Weaver. Must match the container image
processingUnits: 1 # how many Processing Units to use from the license
replicas: 2 # fixed number of replicas (2), as opposed to auto-scale
engfra:
lp: engfra_generic_srv_tnmv_8_4_x_1
processingUnits: 2
scale:
min: 1 # scale down to 1 replica
max: 4 # scale up to 4 replicas on high CPU usage
cpuPercent: 75 # CPU usage (percent) threshold used to scale up or down
# Training Engines are optional
trainingEngines:
replicas: 1