Files
nomad/website/content/docs/job-specification/parameterized.mdx
Aimee Ukasick 986f3c727a Docs: SEO job spec section (#25612)
* action page

* change all page_title fields

* update title

* constraint through migrate pages

* update page title and heading to use sentence case

* fix front matter description

* Apply suggestions from code review

Co-authored-by: Jeff Boruszak <104028618+boruszak@users.noreply.github.com>

---------

Co-authored-by: Jeff Boruszak <104028618+boruszak@users.noreply.github.com>
2025-05-19 09:02:07 -05:00

222 lines
8.0 KiB
Plaintext

---
layout: docs
page_title: parameterized block in the job specification
description: |-
A parameterized job is used to encapsulate a set of work that can be carried
out on various inputs much like a function definition. When the
`parameterized` block is added to a job, the job acts as a function to the
cluster as a whole.
---
# `parameterized` block in the job specification
<Placement groups={['job', 'parameterized']} />
A parameterized job is used to encapsulate a set of work that can be carried out
on various inputs much like a function definition. When the `parameterized`
block is added to a job, the job acts as a function to the cluster as a whole.
The `parameterized` block allows job operators to configure a job that carries
out a particular action, define its resource requirements and configure how
inputs and configuration are retrieved by the tasks within the job.
To invoke a parameterized job, [`nomad job dispatch`][dispatch command] or the equivalent HTTP APIs are
used. When dispatching against a parameterized job, an opaque payload and
metadata may be injected into the job. These inputs to the parameterized job act
like arguments to a function. The job consumes them to change its behavior,
without exposing the implementation details to the caller.
To that end, tasks within the job can add a
[`dispatch_payload`][dispatch_payload] block that
defines where on the filesystem this payload gets written to. An example payload
would be a task's JSON configuration.
Further, certain metadata may be marked as required when dispatching a job so it
can be used to inject configuration directly into a task's arguments using
[interpolation]. An example of this would be to require a run ID key that
could be used to lookup the work the job is suppose to do from a management
service or database.
Each time a job is dispatched, a unique job ID is generated. This
allows a caller to track the status of the job, much like a future or
promise in some programming languages. The dispatched job cannot be
updated after dispatching; to update the job definition you need to
update the parent job.
```hcl
job "docs" {
parameterized {
payload = "required"
meta_required = ["dispatcher_email"]
meta_optional = ["pager_email"]
}
}
```
See the [multiregion] documentation for additional considerations when
dispatching parameterized jobs.
## Requirements
- The job's [scheduler type][batch-type] must be `batch` or `sysbatch`.
## Parameters
- `meta_optional` `(array<string>: nil)` - Specifies the set of metadata keys that
may be provided when dispatching against the job.
- `meta_required` `(array<string>: nil)` - Specifies the set of metadata keys that
must be provided when dispatching against the job.
- `payload` `(string: "optional")` - Specifies the requirement of providing a
payload when dispatching against the parameterized job. The **maximum size of a
`payload` is 16 KiB**. The options for this
field are:
- `"optional"` - A payload is optional when dispatching against the job.
- `"required"` - A payload must be provided when dispatching against the job.
- `"forbidden"` - A payload is forbidden when dispatching against the job.
## Examples
The following examples show non-runnable example parameterized jobs:
### Required inputs
This example shows a parameterized job that requires both a payload and
metadata:
```hcl
job "video-encode" {
# ...
type = "batch"
parameterized {
payload = "required"
meta_required = ["dispatcher_email"]
}
group "encode" {
# ...
task "ffmpeg" {
driver = "exec"
config {
command = "ffmpeg-wrapper"
# When dispatched, the payload is written to a file that is then read by
# the created task upon startup
args = ["-config=${NOMAD_TASK_DIR}/config.json"]
}
dispatch_payload {
file = "config.json"
}
}
}
}
```
### Metadata interpolation
```hcl
job "email-blast" {
# ...
type = "batch"
parameterized {
payload = "forbidden"
meta_required = ["CAMPAIGN_ID"]
}
group "emails" {
# ...
task "emailer" {
driver = "exec"
config {
command = "emailer"
# The campaign ID is interpolated and injected into the task's
# arguments
args = ["-campaign=${NOMAD_META_CAMPAIGN_ID}"]
}
}
}
}
```
### Use periodic with parameterized
Nomad uses an internal hierarchy when scheduling a job that is both `parameterized` and [`periodic`][periodic].
```plaintext
parameterized => periodic => batch
```
Nomad processes a periodic with parameterized job in the following order:
1. Nomad does not dispatch a periodic job with null parameters in the periodic configuration. This forces the parameterized job to take precedence over the periodic job.
2. After Nomad dispatches the parameterized job and gives it parameters, Nomad uses the periodic configuration.
3. Nomad dispatches new jobs according to the periodic configuration that uses thee parameters from the triggering parameterized job.
In this example, the periodic job does not trigger any new jobs
until the operator dispatches the parameterized job at least once. After that, the
dispatched child periodically triggers more children with the given parameters.
```hcl
periodic {
crons = [
"*/40 * * * * * *"
]
}
parameterized {
payload = "required"
meta_required = ["dispatcher_email"]
meta_optional = ["pager_email"]
}
```
There are three columns plus comments in this example output, which is for the preceding periodic, parameterized example job. Scroll to the last column to review the comments.
```
$ nomad job status
ID Type Submit Date
sync batch/periodic/parameterized 2024-11-07T10:43:30+01:00 // Original submitted job
sync/dispatch-1730972650-247c6e97 batch/periodic 2024-11-07T10:44:10+01:00 // First dispatched job with parameters A
sync/dispatch-1730972650-247c6e97/periodic-1730972680 batch 2024-11-07T10:44:40+01:00 // Cron job with parameters A
sync/dispatch-1730972650-247c6e97/periodic-1730972860 batch 2024-11-07T10:47:40+01:00 // Cron job with parameters A
sync/dispatch-1730972760-f79a96e1 batch/periodic 2024-11-07T10:46:00+01:00 // Second dispatched job with parameters B
sync/dispatch-1730972760-f79a96e1/periodic-1730972800 batch 2024-11-07T10:46:40+01:00 // Cron job with parameters B
sync/dispatch-1730972760-f79a96e1/periodic-1730972860 batch 2024-11-07T10:47:40+01:00 // Cron job with parameters B
```
The output illustrates the following:
1. The `sync` job is the parameterized and periodic job. However, Nomad doesn't run the periodic job until you submit with the required parameters.
2. The `sync/dispatch-1730972650-247c6e97` job contains parameters and triggers the batch jobs, named `sync/dispatch-1730972650-247c6e97/periodic-1730972680` and `sync/dispatch-1730972650-247c6e97/periodic-1730972860`.
3. The `sync/dispatch-1730972760-f79a96e1` job assigns new parameters and triggers new batch jobs that use those parameters.
If you need to force the periodic job, force the the corresponding parameterized job.
This example forces the first dispatched job with parameters A from the preceding example.
```
$ nomad job periodic force sync/dispatch-1730972650-247c6e97
```
[batch-type]: /nomad/docs/job-specification/job#type 'Batch scheduler type'
[dispatch command]: /nomad/docs/commands/job/dispatch 'Nomad Job Dispatch Command'
[resources]: /nomad/docs/job-specification/resources 'Nomad resources Job Specification'
[interpolation]: /nomad/docs/runtime/interpolation 'Nomad Runtime Interpolation'
[dispatch_payload]: /nomad/docs/job-specification/dispatch_payload 'Nomad dispatch_payload Job Specification'
[multiregion]: /nomad/docs/job-specification/multiregion#parameterized-dispatch
[periodic]: /nomad/docs/job-specification/periodic