Hello,
I’m using the Orchestrator CE APIs. This is my request to Start a job. I’ve got the Release Key of the Process that I need to run as a job via the /Releases API, and the Robot ID using the /Sessions API call.
Further, the Robot Role has Create+Edit permissions on the Jobs, and View Permissions on Process and Environment.
However, I’m constantly getting 400 (Bad request) errors. Please see my API request below and let me know what’s wrong in this request.
You are almost there. But instead of using the /Sessions endpoint which returns the ID of the session, you should get the Robot ID from the /Robots endpoint.
I switched over to the /Robot endpoint . Yes, this time it’s giving me a different Robot Id.
That said, I'm back to square 1 with the 400 Bad Request error after making all these changes. The response body is completely empty!
The result is the same even after I repeat the test between two completely different Orchestrator instances. Every other API call up to this point is working except this one.
I hate to post these many updates for such a simple issue. Something is really flaky here.
I came back home and repeated the experiment that had failed earlier. And it worked perfectly! The job kicked off and successfully completed.
That’s the good news!
Ten minutes later I repeated the whole series of steps and this time, it crashed and burned with a 400 error!
To my defense, the test scripts have been saved in Talend as a project. It’s not that I’m manually typing in these API requests every time.
Is there some limit in the CE version that the StartJob API cannot be called twice in under 15 minutes?
Yes, I made a couple more attempts and there was no change.
One significant observation:
Once the 400 error is hit, no matter how many times you try, the API end point does not recover from it. Even submitting a brand new request with a newly refreshed token and trying to start a different job has no impact. It still throws a 400 error.
I feel something is wrong with this endpoint because all others work consistently without any issues.
I gave this a shot 2 days after my last failure attempt and it’s working. I used the exact same API call scripts, with a newly acquired token. I ran 3 tests. Started the same job twice within the first 5 minutes.
And then after a few minutes, I got the release key for a second job and started it via the API.
All three tests succeeded and the jobs ran successfully. I feel that the StartJob end point was in flux over the past few days I've been struggling with this. That said, I will run these tests for a different Orchestrator CE instance from a different Robot machine so that I can confirm consistency and reliability.
Test results with timestamps of the current tests below:
And indeed, the tests after an hour have magnificently failed! The only differences was an addition to the JSON to pass in arguments to the job I was testing.
Once the test failed, I tried running the job under default state with the original JSON, but despite refreshing the token the job did not run. And just like 2 days ago, the StartJob API endpoint seem to be paralyzed by a 400 error and has become unresponsive !
I was able to replicate a similar behavior using Postman and doing some changes in the request. The problem seems to be the invalid forgery token sent by the app. After restarting Postman the issue was solved. I am not familiar with Talend but it should allow cookie or CSRF token regeneration.
This article exemplifies accessing Google APIs using “Client ID” and “Client Secret”. But isn’t that what we do each time we refresh tokens for the CE APIs? Fetch the Client ID from the API Access dialog and use that to get a new token?
Or is this article suggesting that we do something more?
Using the credentials from API Access and POST https://account.uipath.com/oauth/token you should retrieve the access token (valid 24h). You do not have to do this for every call, you only need to retrieve the access_token when it expires.
Can you try to use postman instead ? At least we isolate the problem (Talend or Orchestrator CE).
I have shared an example so all you have to do is to populate the variables for oauth/token with your credentials (taken from cloud portal - API Access) and update the StartJobs Body with your IDs.
Yes, but I tried to refresh my token so that I could repeat the test from scratch once I ran into the 400 error. Otherwise it would be every 24 hours. FYI: We have other non-RPA Cloud integrations where tokens expire every hour and we have to check for token expiry before we make the actual call for data.
I tried to avoid installing yet another executable on my laptop, but it looks like that’s unavoidable. I will experiment with postman and keep you posted.
Side bar question: I see that the User-Key changes each time I bring up the API Access dialog. How do we manage this every 24 hours? Do we have to write another process to acquire the User-Key in an automation fashion before we make a call to refresh the token?
Postman is very light and has a good reputation. Also is fast and easy to test API calls.
You do not have to change the user-key. Once you have it generated you can use it until you revoke access (for all previous generated keys). This can be seen only once as a security measure and the next time you request API access a new key is generated (but the old one is still valid). Also the user-key should be treated as a password and kept secret.
Before I observed the User Key changing, I was planning to store the User-Key in the WCM of the Robot machine as a quasi-password. Meaning, the user name would be the logical Account+Logical name and the password would be the User Key. I would retrieve the User Key by providing a credential name during run time for the purposes of refreshing the token, or for normal use.
This way, the User Key would not be a hard-coded artifact, and the application would seamlessly move from DEV to UAT/PROD.
Here are the test results. The Creation times published below are the responses received to the StartJob API Requests.
Summary: Over the Span of 3 hours, 4 tests were conducted. Postman tests succeeded whilst the exact same test (and only test) repeated in Talend failed.
Postman, StartJob Test #1 >> Success!
“CreationTime”: “2019-12-25T18:50:12.4182491Z”,
Postman, StartJob Test #2 - Same token, different Job, 15 minutes later >> Success
“CreationTime”: “2019-12-25T19:06:28.2694346Z”,
Talend, StartJob Test #1 after successful Postman tests #1 and #2 above
Job #2 was repeated exactly as above, but refreshing a token, 4 minutes later
date: Wed, 25 Dec 2019 19:10:04 GMT
FAILED: 400 Error
Back to Postman 3 hours later, StartJob, Repeat Test #2 exactly without any changes! >> Success!
“CreationTime”: “2019-12-25T22:01:06.569073Z”
I’m not sure now if Talend is the problem, aren’t the APIs supposed to work client-agnostic?