Occasionally "Operation not permitted"

[Kain-90]

I have read the issue

But my problem is different, with the same parameters, sometimes my code execution block can be successful, but sometimes it reports an error: Operation not allowed

I am also unable to find the relevant execution logs, and I have encountered the same problem while trying to upgrade the version. I have not found a stable reproduce logic, it interrupts the workflow execution every time it appears.

BTW, I'm using a self-hosted deployment method, and the version is 0.6.16.

The simplest step for reproduction:

1. Create the example code execution block
2. Rapidly click "Run this step" multiple times (about 10 to 30 times)

[law52525]

I have the same situation. Whether I use the newest code or the code of version 0.2.10, this problem sometimes happens. And I notice that when there are a lot of concurrent requests, this problem is very likely to appear.

[18827555809]

I have the same situation. Whether I use the newest code or the code of version 0.2.10, this problem sometimes happens. And I notice that when there are a lot of concurrent requests, this problem is very likely to appear.

I have the same situation，When concurrency is particularly high, this problem may occur. I don't know if it's related to the system kernel

[18827555809]

I have read the issue

But my problem is different, with the same parameters, sometimes my code execution block can be successful, but sometimes it reports an error: Operation not allowed

I am also unable to find the relevant execution logs, and I have encountered the same problem while trying to upgrade the version. I have not found a stable reproduce logic, it interrupts the workflow execution every time it appears.

BTW, I'm using a self-hosted deployment method, and the version is 0.6.16.

The simplest step for reproduction:

1. Create the example code execution block
2. Rapidly click "Run this step" multiple times (about 10 to 30 times)

I have already solved this problem. The reason is that in concurrent scenarios, there may be user mode processes that actively release the CPU. This process requires a system call sched_yield because there are many waiting processes, and the operating system will choose to execute high priority jobs. At this time, low priority jobs need to actively release the CPU, but we do not give sched_yield permission. Therefore, when encountering the process of actively releasing the CPU, an error will be reported: bad system call， In my Linux kernel, it is 24

[Kain-90]

I have read the issue
But my problem is different, with the same parameters, sometimes my code execution block can be successful, but sometimes it reports an error: Operation not allowed
I am also unable to find the relevant execution logs, and I have encountered the same problem while trying to upgrade the version. I have not found a stable reproduce logic, it interrupts the workflow execution every time it appears.
BTW, I'm using a self-hosted deployment method, and the version is 0.6.16.
The simplest step for reproduction:

1. Create the example code execution block
2. Rapidly click "Run this step" multiple times (about 10 to 30 times)

I have already solved this problem. The reason is that in concurrent scenarios, there may be user mode processes that actively release the CPU. This process requires a system call sched_yield because there are many waiting processes, and the operating system will choose to execute high priority jobs. At this time, low priority jobs need to actively release the CPU, but we do not give sched_yield permission. Therefore, when encountering the process of actively releasing the CPU, an error will be reported: bad system call， In my Linux kernel, it is 24

It makes sense logically, but in practice, I don't really have any concurrency scenarios since I'm manually clicking the 'start' button, and as you know, it's physically impossible to click multiple times at the exact same moment.