Hello, I am having an issue with the timing of my jobs starting.
I am using a ServiceNow connector to submit a queue item to a queue in UiPath, which then triggers a job to start.
This piece works fine, and the time from ServiceNow submitting the request, to the queue item appearing in New status in the queue is generally less than two seconds.
The problem I’m seeing, however, is a long and steadily-increasing delay from the time that item appears in the queue, until the time the job actually starts.
I can monitor the queue directly and see the queue item arrive in New status, and then sit for up to 45 seconds before switching to In Progress, indicating the delay isn’t in the process itself since it is happening before execution even starts.
I’ve pulled data from ServiceNow (see the attachment), where we log the time the request was sent to UiPath and the time UiPath actually started executing the job.
u_uipath_request.xlsx (28.4 KB)
When we first put this process into production about two months ago, the average delay over our first week was around 19 seconds, but it is now over 25 seconds with many jobs taking nearly a minute to begin execution.
A few other relevant facts -
- This is our only production process running on a single runtime, and no other jobs are waiting to complete before these new requests start
- We are running the robot on an AWS EC2 T3 extra-large instance which meets all of the recommended specs for the UiPath Robot
–All CPU usage, memory usage, and hard drive space on the machine are nominal
- Our Orchestrator implementation is via Automation Cloud
Since this problem is clearly cumulative, and in the two months this process has been live we’ve seen such a significant increase in delay already, we are obviously concerned about this continuing to grow and becoming untenable; this process is a time-sensitive, customer-facing robot that needs to provide results in around 90 seconds and these delays already represent a significant chunk of that time budget.
Has anyone seen a similar issue in their implementations, or have any advice on how we can remediate this? Thank you!