verbo.sokka — ai.verbo/sokka 0.0.1-SNAPSHOT

Background task management library for Clojure.

create-task! clj

(create-task! taskq task)

Create and returns a new task.

failed clj

(failed)

(failed {:keys [error-message]})

list-tasks clj

(list-tasks taskq
            {:keys [from to topic sub-topic] :as filters}
            {:keys [limit] :as cursor})

List all tasks for a given topic.

ok clj

(ok)

snoozed clj

(snoozed)

(snoozed {:keys [snooze-time]})

task clj

(task taskq task-id)

Get task by task-id. Returns nil when no task is found.

taskq clj multimethod

tasks clj

(tasks taskq task-group-id)

Get tasks for a given task-group-id. Returns nil when no tasks are found for the supplied input.

worker clj

(worker {:keys [taskq topic pid pfn lease-time-ms keepalive-ms timeout-ms
                max-poll-interval-ms]
         :as opts})

Polls the task service on the given topic for a task and when available, obtains a lease for the task, and calls processor function pfn in a separate thread, passing the task as an argument. It also spins up a sidecar thread (keepalive) to extend the lease of the task periodically, while pfn is being executed.

processor-fn - should accept task as an argument, perform thecleanup-leased-tasks operation corresponding to the task and return a tuple containing [event-name opts]. valid event-names and args are: [:sokka/completed nil], [:sokka/failed, {:keys [error-message]}] [:sokka/snoozed, {:keys [snooze-time]}].

; keepalive-ms - keepalive timeout, will be lease-time * 0.7 ; task-timeout - how long do we expect the task to run. this can be overridden at the task level, but set at the worker level

The worker will wait until worker-timeout-ms has passed for the task to complete, if the task isn't complete by then, it closes the task, acknowledges the task with status = :failed and carries on. If the task completes successfully, it acknowledges the task with status = :ok and carries on.

The worker polls for tasks using an exponentially increasing sleeper function to prevent the worker from flooding the queue with requests during inactivity.