Code-server instances "stay open" after being closed from browser

[Sharpz7]

OS/Web Information

• Web Browser: Chrome
• Local OS: Windows
• Remote OS: Docker Linux
• Remote Architecture: Docker Compose Contianer
• code-server --version:

Steps to Reproduce

1. Open htop
2. Open multiple tabs and close them again when code-server loads
3. The instance stays open, with HTOP showing that the CPU usuage and RAM increases with each closed instance until it crashes.

Expected

The RAM and CPU usage should decrease after the instance is closed

Actual

It doesn't :)

Logs

https://files.mcaq.me/8xad.png
I also don't understand why the code-server items are tied to my user "adam" instead of root. This doesn't make much sense to me?

This is difficult to show with logs, or at least I don't know how to do it well with my setup.

[geiseri]

This is 50% by design but also related to #3947. The idea is that you can keep a session alive, but you should be able to close the instance when you close the folder.

[code-asher]

Does the reproduction require opening a terminal or does this happen when opening code-server without a terminal open?

[Sharpz7]

After further testing, it seems to be based on a timeout?

So after x mins, the instances close.

[code-asher]

Ahh yup the timeout is in case the connection drops so the browser can reconnect. If there is no reconnection within the timeout the terminal instance is killed.

[Sharpz7]

Does this issue extend to each reload of the page? I reloaded my page 3 times and the RAM usage went to 8GB with the SWAP being maxed to. maxed out my 4 cores as well and crashed all my docker containers.

I then tried removing a bunch of my extensions, bracketizers, spelling checkers, CSS navigators and tabnine. I reckon whats happened is one of them was running in the background and then not closing again? I don't have time for a full testing session for all the extensions (tbh I don't need them) but I could provide a list for you guys?

https://files.mcaq.me/mvrt.png
My usage has now dropped significantly

[code-asher]

That sounds like a bug to me. It should be reconnecting existing instances but it sounds like we have a leak.

[Sharpz7]

My bet with such high RAM usage is that it was probably tabnine. I would experiment with it but it takes my server down so often it's not very ideal hahaha

…

On Mon, 25 Oct 2021, 20:02 Asher, ***@***.***> wrote: That sounds like a bug to me. It should be reconnecting existing instances but it sounds like we have a leak. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#4351 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALCTWHUG3SWSHHXJ6O2ICPTUIWST5ANCNFSM5F56W3ZA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[code-asher]

Ah yeah that makes sense. Each refresh spawns a new extension host but the old one is supposed to be killed immediately. Maybe they are not getting killed...? They are labeled with `--type extensionHost` I think so you could check the process listing and see if there is more than one.

[Sharpz7]

In that case, give me a few days and I'll attempt to check Thanks

…

On Mon, 25 Oct 2021, 20:35 Asher, ***@***.***> wrote: Ah yeah that makes sense. Each refresh spawns a new extension host but the old one is supposed to be killed immediately. Maybe they are not getting killed...? They are labeled with `--type extensionHost` I think so you could check the process listing and see if there is more than one. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#4351 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALCTWHSOF3VT3NPHDONW72DUIWWO7ANCNFSM5F56W3ZA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[Sharpz7]

Alright. So since upgrading to 3.12.0 I have not had this issue. There are two reasons why that could be the case:

• Tabnine wasn't the issue
• 3.11.1 was the issue (I think this is likely, I had tabnine in versions before this with no issues

Regardless, I do not have the time to test this completely, especially after I have just got 3.12.0 working.

Thanks for all the help