Repository navigation
Understanding the limits of how Active jobs vs. Queued jobs work #45
Replies: 6 comments
0 replies
0 replies
0 replies
|
I'm so glad you were able to make it work! Yeah, I think there are a number of places where a ~1 second delay could be being introduced during a workflow. I'll look into this deeper and see if I can track it down. |
0 replies
|
I just released xyOps v0.9.27, which should fix this issue. When a job completes, it now "ticks" the job queue (rather than waiting for the next clock second), which should immediately launch another one if a slot opened up. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment






Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Hi!
I'm still researching my use case tests and I don't quite understand the management of event execution limits and the global configuration of xyops.
I have the event configured as follows

According to the global configuration, the max_jobs_per_min parameter has a default value of 100.
In my tests, I am send 200 jobs from an event in a loop without any sleep, so in 3-4 seconds I have already sent the 200 run_event requests. The first 100 requests are queued and executed concurrently with a maximum of 10 according to the event parameters.
The thing is, for the next 100 requests, I get an HTTP 500 response: {“code”:“event”,‘description’:“Failed to launch job: Global jobs per minute limit has been reached (100/min)”}. This is actually correct for the max_jobs_per_min parameter, but does this behavior make sense?
So, I'm wondering if it's necessary to set the limit on all executions or only on those that are active? What I mean by this is that if I send the execution request to an event that has the queue activated and also concurrent executions, it should accept the registration but put them in job.state = queued status. In this case, I don't reach the limit of 100, but if I only have 10 concurrent executions, I still have “90 slots” free.
I would also like to add that it is possible to estimate the max_jobs_per_min limit in some way based on the number of cores/vcpu/ram.
All reactions