Skip to content

README.md

Latest commit

 

History

History
131 lines (103 loc) · 6.62 KB

README.md

File metadata and controls

131 lines (103 loc) · 6.62 KB

JetStream for OpenFaaS queue-worker

Use this chart to deploy additional queues and queue-workers, for functions which need to be separated out from the default queue.

Note: This chart requires OpenFaaS for Enterprises. Please reach out to us to discuss upgrading your installation.

Reasons why you may want an additional queue-worker:

  • You run workloads for your own product, and for tenants, and want to separate the two
  • You want to run some queues with different settings, such as maxInflight (perhaps a function has a specific amount of requests it can handle at once), upstreamTimeout (longer timeouts), mode (static/function based consumers)
  • Certain long-running invocations may "clog up" the queue and starve requests from other functions, so you want to run those separately

Install an additional queue-worker

helm upgrade --install \
  slow-fns openfaas/queue-worker \
  --namespace openfaas \
  --set maxInflight=5 \
  --set mode=static \
  --set nats.stream.name=slow-fns \
  --set nats.consumer.durableName=slow-fns-workers \
  --set upstreamTimeout=15m

It's recommended to take the NAME of the queue i.e. slow-fns and then use it as prefix in the following way for the above configuration:

  • nats.stream.name - NAME-requests
  • nats.consumer.durableName - NAME-workers

As an alternative to using --set, you could also write your own YAML file. Below is the equivalent configuration in a values.yaml file, for instance values-slow-fns.yaml:

maxInflight: 5
mode: static
nats:
  stream:
    name: slow-fns
  consumer:
    durableName: slow-fns-workers
upstreamTimeout: 15m

Then pass -f ./values-slow-fns.yaml to the helm upgrade --install command instead of the --set flags.

The chart will append a suffix of -queue-worker to the release name given above, so the name of the queue-worker will be slow-fns-queue-worker.

Upon start-up, the queue-worker will create a NATS JetStream stream named slow-fns. Depending on the queue mode either one static consumer will be created, or if the mode is set to function, a consumer will be created for each function that has invocations pending.

Remember to set the com.openfaas.queue annotation for your functions, so that their requests get submitted to the correct queue (NATS JetStream Stream).

For example:

# Runs on the slow queue
faas-cli store deploy sleep --name slow-fn --annotation com.openfaas.queue=slow-fns

faas-cli invoke --async slow-fn <<< ""

# Runs on the default queue
faas-cli store deploy env --name fast-fn

faas-cli invoke --async fast-fn <<< ""

To remove a queue-worker, run:

helm uninstall --namespace openfaas \
  slow-fns

Install a development version of the chart

The below instructions are for development on the chart itself, or if you want to test changes to the chart's templates that have not yet been released.

Checkout the code, and cd to the folder:

git clone https://github.com/openfaas/faas-netes --depth=1
cd faas-netes/chart/queue-worker
helm upgrade --install \
  slow-fns ./ \
  --namespace openfaas \
  --set maxInflight=5 \
  --set mode=static \
  --set nats.stream.name=slow-fns \
  --set nats.consumer.durableName=slow-fns-workers \
  --set upstreamTimeout=15m

Configuration

Parameter Description Default
image The queue-worker image that should be deployed See values.yaml
replicas Number of queue-worker replicas to create 1
queueName Name of the queue if you want it to be different to the stream name "" - when empty, defaults to nats.stream.name
mode Queue operation mode: static (OpenFaaS Standard) or function (requires OpenFaaS for Enterprises) static
maxInflight Control the concurrent invocations 1
adaptiveConcurrency Enable adaptive concurrency limiting for functions. This setting only takes effect when mode is set to function. true
queuePartitions Number of queue partitions 1
partition Queue partition number this queue should subscribe to 0
consumer.inactiveThreshold If a function is inactive (has no invocations) for longer than this threshold its consumer will be removed to save resources 30s
consumer.pullMaxMessages PullMaxMessages limits the number of messages to be buffered per consumer. Leave empty to use optimized default for the selected queue mode ``
upstreamTimeout Maximum duration of upstream function call 1m
maxRetryAttempts The amount of times to try sending a message to a function before discarding it 10
maxRetryWait The maximum amount of time to wait between retries 120s
initialRetryWait The amount of time to wait for the first retry 10s
httpRetryCodes A comma-separated list of HTTP status codes which the queue worker will retry when received from a function 408,429,500,502,503,504
backoff The backoff algorithm used for retries. Must be one off exponential, full or equal exponential
gateway.host The host at which the OpenFaaS gateway can be reached http://gateway.openfaas
gateway.port The port at which the OpenFaaS gateway can be reached 8080
insecureTLS Enable insecure tls for callbacks false
nats.host The host at which the NATS JetStream server can be reached nats.openfaas
nats.port The port at which the NATS JetStream server can be reached 4222
nats.stream.name Name of the NATS JetStream stream to use faas-request
nats.stream.replicas Number of JetStream stream replicas to create 1
nats.consumer.durableName The name of the NATS JetStream consumer to use faas-workers
nats.consumer.ackWait AckWait configures how long the NATS waits for an acknowledgement before redelivering the message 30s
logs.debug Print debug logs false
logs.format The log encoding format. Supported values: json or console console
resources.requests.memory Memory resource request 120Mi
resources.requests.cpu CPU resource request 50m
imagePullPolicy Image pull policy IfNotPresent
nodeSelector NodeSelector for deployment {}
tolerations Tolerations for deployment []
affinity Aaffinity rules for deployment {}