|
| 1 | +--- |
| 2 | +short_description: How to rerun a step execution in Node.js |
| 3 | +thumbnail: https://res.cloudinary.com/pipedreamin/image/upload/v1646841376/docs/icons/icons8-time-96_kupxpi.png |
| 4 | +--- |
| 5 | + |
| 6 | +# Pause, resume, and rerun a workflow |
| 7 | + |
| 8 | +You can use `$.flow.suspend` and `$.flow.rerun` to pause a workflow and resume it later. |
| 9 | + |
| 10 | +This is useful when you want to: |
| 11 | + |
| 12 | +- Pause a workflow until someone manually approves it |
| 13 | +- Poll an external API until some job completes, and proceed with the workflow when it's done |
| 14 | +- Trigger an external API to start a job, pause the workflow, and resume it when the external API sends an HTTP callback |
| 15 | + |
| 16 | +We'll cover all of these examples below. |
| 17 | + |
| 18 | +[[toc]] |
| 19 | + |
| 20 | +## `$.flow.suspend` |
| 21 | + |
| 22 | +Use `$.flow.suspend` when you want to pause a workflow and proceed with the remaining steps only when manually approved or cancelled. |
| 23 | + |
| 24 | +For example, you can suspend a workflow and send yourself a link to manually resume or cancel the rest of the workflow: |
| 25 | + |
| 26 | +```javascript |
| 27 | +export default defineComponent({ |
| 28 | + async run({ $ }) { |
| 29 | + const { resume_url, cancel_url } = $.flow.suspend() |
| 30 | + $.send.email({ |
| 31 | + subject: "Please approve this important workflow", |
| 32 | + text: `Click here to approve the workflow: ${resume_url}, and cancel here: ${cancel_url}`, |
| 33 | + }) |
| 34 | + // Pipedream suspends your workflow at the end of the step |
| 35 | + }, |
| 36 | +}) |
| 37 | +``` |
| 38 | + |
| 39 | +You'll receive an email like this: |
| 40 | + |
| 41 | +<div> |
| 42 | + <img src="https://res.cloudinary.com/pipedreamin/image/upload/v1655272047/docs/approve-workflow_oc06k3.png" alt="approve this workflow" width="100%"/> |
| 43 | +</div> |
| 44 | + |
| 45 | +And can resume or cancel the rest of the workflow by clicking on the appropriate link. |
| 46 | + |
| 47 | +### `resume_url` and `cancel_url` |
| 48 | + |
| 49 | +In general, calling `$.flow.suspend` returns a `cancel_url` and `resume_url` that lets you cancel or resume paused executions. Since Pipedream pauses your workflow at the _end_ of the step, you can pass these URLs to any external service before the workflow pauses. If that service accepts a callback URL, it can trigger the `resume_url` when its work is complete. |
| 50 | + |
| 51 | +These URLs are specific to a single execution of your workflow. While the workflow is paused, you can load these in your browser or send any HTTP request to them: |
| 52 | + |
| 53 | +- Sending an HTTP request to the `cancel_url` will cancel that execution |
| 54 | +- Sending an HTTP request to the `resume_url` will resume that execution |
| 55 | + |
| 56 | +If you resume a workflow, any data sent in the HTTP request is passed to the workflow and returned in the `$resume_data` [step export](/workflows/steps/#step-exports) of the suspended step. For example, if you call `$.flow.suspend` within a step named `code`, the `$resume_data` export should contain the data sent in the `resume_url` request: |
| 57 | + |
| 58 | +<div> |
| 59 | + <img src="https://res.cloudinary.com/pipedreamin/image/upload/v1655271815/docs/resume_data_lafhxr.png" alt="resume data step export" width="350px"/> |
| 60 | +</div> |
| 61 | + |
| 62 | +### Default timeout of 24 hours |
| 63 | + |
| 64 | +By default, `$.flow.suspend` will automatically resume the workflow after 24 hours. You can set your own timeout (in milliseconds) as the first argument: |
| 65 | + |
| 66 | +```javascript |
| 67 | +export default defineComponent({ |
| 68 | + async run({ $ }) { |
| 69 | + // 7 days |
| 70 | + const TIMEOUT = 1000 * 60 * 60 * 24 * 7 |
| 71 | + $.flow.suspend(TIMEOUT) |
| 72 | + }, |
| 73 | +}) |
| 74 | +``` |
| 75 | + |
| 76 | +## `$.flow.rerun` |
| 77 | + |
| 78 | +Use `$.flow.rerun` when you want to run a specific step of a workflow multiple times. This is useful when you need to start a job in an external API and poll for its completion, or have the service call back to the step and let you process the HTTP request within the step. |
| 79 | + |
| 80 | +### Polling for the status of an external job |
| 81 | + |
| 82 | +Sometimes you need to poll for the status of an external job until it completes. `$.flow.rerun` lets you rerun a specific step multiple times: |
| 83 | + |
| 84 | +```javascript |
| 85 | +import axios from 'axios' |
| 86 | + |
| 87 | +export default defineComponent({ |
| 88 | + async run({ $ }) { |
| 89 | + const MAX_RETRIES = 3 |
| 90 | + // 10 seconds |
| 91 | + const DELAY = 1000 * 10 |
| 92 | + const { run } = $.context |
| 93 | + // $.context.run.runs starts at 1 and increments when the step is rerun |
| 94 | + if (run.runs === 1) { |
| 95 | + // $.flow.rerun(delay, context (discussed below), max retries) |
| 96 | + $.flow.rerun(DELAY, null, MAX_RETRIES) |
| 97 | + } |
| 98 | + else if (run.runs === MAX_RETRIES + 1) { |
| 99 | + throw new Error("Max retries exceeded") |
| 100 | + } |
| 101 | + else { |
| 102 | + // Poll external API for status |
| 103 | + const { data } = await axios({ |
| 104 | + method: "GET", |
| 105 | + url: "https://example.com/status" |
| 106 | + }) |
| 107 | + // If we're done, continue with the rest of the workflow |
| 108 | + if (data.status === "DONE") return data |
| 109 | + |
| 110 | + // Else retry later |
| 111 | + $.flow.rerun(DELAY, null, MAX_RETRIES) |
| 112 | + } |
| 113 | + }, |
| 114 | +}) |
| 115 | +``` |
| 116 | + |
| 117 | +`$.flow.rerun` accepts the following arguments: |
| 118 | + |
| 119 | +```javascript |
| 120 | +$.flow.rerun( |
| 121 | + delay, // The number of milliseconds until the step will be rerun |
| 122 | + context, // JSON-serializable data you need to pass between runs |
| 123 | + maxRetries, // The total number of times the step will rerun. Defaults to 10 |
| 124 | +) |
| 125 | +``` |
| 126 | + |
| 127 | +### Accept an HTTP callback from an external service |
| 128 | + |
| 129 | +When you trigger a job in an external service, and that service can send back data in an HTTP callback to Pipedream, you can process that data within the same step using `$.flow.retry`: |
| 130 | + |
| 131 | +```javascript |
| 132 | +import axios from 'axios' |
| 133 | + |
| 134 | +export default defineComponent({ |
| 135 | + async run({ steps, $ }) { |
| 136 | + const TIMEOUT = 86400 * 1000 |
| 137 | + const { run } = $.context |
| 138 | + // $.context.run.runs starts at 1 and increments when the step is rerun |
| 139 | + if (run.runs === 1) { |
| 140 | + const { cancel_url, resume_url } = $.flow.rerun(TIMEOUT, null, 1) |
| 141 | + |
| 142 | + // Send resume_url to external service |
| 143 | + await axios({ |
| 144 | + method: "POST", |
| 145 | + url: "your callback URL", |
| 146 | + data: { |
| 147 | + resume_url, |
| 148 | + cancel_url, |
| 149 | + } |
| 150 | + }) |
| 151 | + } |
| 152 | + else if (run.runs === 2) { |
| 153 | + throw new Error("External service never completed job") |
| 154 | + } |
| 155 | + // When the external service calls back into the resume_url, you have access to |
| 156 | + // the callback data within $.context.run.callback_request |
| 157 | + else { |
| 158 | + const { callback_request } = run |
| 159 | + return callback_request |
| 160 | + } |
| 161 | + }, |
| 162 | +}) |
| 163 | +``` |
| 164 | + |
| 165 | +### Passing `context` to `$.flow.rerun` |
| 166 | + |
| 167 | +Within a Node.js code step, `$.context.run.context` contains the `context` passed from the prior call to `rerun`. This lets you pass data from one run to another. For example, if you call: |
| 168 | + |
| 169 | +```javascript |
| 170 | +$.flow.rerun(1000, { hello: "world" }) |
| 171 | +``` |
| 172 | + |
| 173 | +`$.context.run.context` will contain: |
| 174 | + |
| 175 | +<div> |
| 176 | + <img src="https://res.cloudinary.com/pipedreamin/image/upload/v1655274732/docs/Screen_Shot_2022-06-14_at_11.32.06_PM_dmzgkh.png" alt="resume data step export" width="250px"/> |
| 177 | +</div> |
| 178 | + |
| 179 | +### `maxRetries` |
| 180 | + |
| 181 | +By default, `maxRetries` is **10**. |
| 182 | + |
| 183 | +When you exceed `maxRetries`, the workflow proceeds to the next step. If you need to handle this case with an exception, `throw` an error from the step: |
| 184 | + |
| 185 | +```javascript |
| 186 | +export default defineComponent({ |
| 187 | + async run({ $ }) { |
| 188 | + const MAX_RETRIES = 3 |
| 189 | + const { run } = $.context |
| 190 | + if (run.runs === 1) { |
| 191 | + $.flow.rerun(1000, null, MAX_RETRIES) |
| 192 | + } |
| 193 | + else if (run.runs === MAX_RETRIES + 1) { |
| 194 | + throw new Error("Max retries exceeded") |
| 195 | + } |
| 196 | + }, |
| 197 | +}) |
| 198 | +``` |
| 199 | + |
| 200 | +## Behavior when testing |
| 201 | + |
| 202 | +When you're building a workflow and test a step with `$.flow.suspend` or `$.flow.rerun`, it will not suspend the workflow, and you'll see a message like the following: |
| 203 | + |
| 204 | +> Workflow execution canceled — this may be due to `$.flow.suspend()` usage (not supported in test) |
| 205 | +
|
| 206 | +These functions will only suspend and resume when run in production. |
| 207 | + |
| 208 | +## Invocations when using `suspend` / `rerun` |
| 209 | + |
| 210 | +Each time workflows are resumed, Pipedream charges [an invocation](/pricing/#invocations). For example, when you call `$.flow.suspend`, you're charged an invocation for the initial event data, and another invocation when you resume the request. |
0 commit comments