Windows: Slow service start (fluent-package) #618
Description
Abst
- Windows service startup is noticeably slower for fluent-package
- This problem can cause timeout errors (EventID 7000 or EventID 7009)
- This problem occurs in environments with multiple CPUs
- This is the problem of win32-service with Ruby 3 series being slow
Details
Windows service startup is noticeably slower for fluent-package.
I'm talking about the time to reach the Running state. In other words, the time to finish StartServiceCtrlDispatcher.
For td-agent, it takes about 5 seconds, but for fluent-package it can take more than 20 seconds.
Windows service needs to finish StartServiceCtrlDispatcher in 30 seconds by default.
So, this problem can cause timeout errors.
EventID 7000
The testservice service failed to start due to the following error:
The service did not respond to the start or control request in a timely fashion.
EventID 7009
A timeout was reached (30000 milliseconds) while waiting for the testservice service to connect.
Reproduce
If you try to start the service several times, it will often take 10 to 20 seconds.
I confirm this problem in the following:
- Windows 10 Home 64bit
- 4 core CPU, 8 logical processors (Intel(R) Core(TM) i7-8565U CPU @ 1.80GHz 1.99GHz)
- 16GB RAM
- AWS EC2 instance Windows Server 2019 Base
- t2.medium
- t2.large
- c5.large
Interestingly, this problem does not occur in t2.small EC2 instance type.
I think it occurs in environments with multiple CPUs.