Process killed when live-reload quarkus Java application

[vanduc2514]

Process killed when live-reload quarkus Java application

[aviramha]

It could be that we're attached to the debugger as well, making it a bit slower.
I'd try to use the extension to see if it helps.

[aviramha]

Thanks for reporting this! This seems to happen because of idle_timeout. The internal proxy component (part that runs on your machine) aggregates "layer" (the part that loads into your local process) connections. When there are no connections for a said duration, it cleans up. I assume the issue is that sometimes the reload time > idle_timeout which means that until the new connection is established the internal proxy already dies.
You can control that timeout by setting

"internal_proxy": {"idle_timeout": 10}

The timeout is in seconds, and by default it is 5 seconds. Can you try increasing it and seeing if it happens again?

[vanduc2514]

Actually I increased the idle_timeout to 1200 (20 minutes) but the issue can still be reproduced. It usually happens if I have a compilation error in my VSCode IDE when writing some code. I use autosave feature of VSCode and the IDE compile the code everytime the file is saved automatically.

Also, the performance in VSCode is greatly impact, much slower, compare to using mirrord exec bash then run the Quarkus application in the terminal. So I guess because Quarkus is executed using launch configuration in VSCode, so the application is executed by mirrord but from a process that VSCode created. VSCode can control mirrord from its spawned process, hence it can kill mirrord execution.

I haven't tried with manul saving (manual compilation) in VSCode but I doubt it will work, since there is less synchronization happening.

[aviramha]

Can you attach the same log as above reproduced with the new idle timeout please?

[vanduc2514]

Sure, this is the trace log

internal_proxy.log

And the configuration is

"internal_proxy": {
        "log_destination": "/tmp/internal_proxy.log",
        "log_level": "mirrord=trace",
        "idle_timeout": 1200
}

[aviramha]

I see it fails on accessing /workspace because it tries to read it from remote.
Can you try to add it to local path?

{
  "feature": {
    "fs": {
      "mode": "read",
      "local": ["/workspace.+"]
    }
  }
}

[vanduc2514]

I tried with the above configuration, Quarkus process did not get killed this time but the performance of VSCode is much slower or freeze for code completion, suggestion, compare to not using that configuration. Sometimes the VSCode debugger is even terminated while I write some code.

Increasing the memory of Java language server helped with the slowing / freezing of VSCode (not entirely) but I don't need to do that if I use mirrord to open a new shell and run Quarkus in development mode within that shell.

Do you think using the mirrord VSCode extension can help with this issue ?