You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After update to 0.27.2, I'm getting "Node.js backend failed to start / Error: Plugin resources not ready". Worked fine in 0.27.1 though. Claude desktop works fine as well.
Yeah... it's kind of weird that it only gets fixed after a reboot.
If a reboot is what fixes it, something must survive closing the IDE.
My guess is that a backend process from the older version is still alive and holding on to the plugin files.
If that's what happens, the plugin should be able to survive it on its own, so that's the direction I'm going to fix.
There is just one thing I'd like to ask you to check.
Next time it happens, close CLion completely, open Task Manager, and see if a node.exe is still there.
If you keep your project inside WSL, it will show up as wsl.exe instead of node.exe.
If it is there, end just that process and start CLion again without rebooting, and let me know whether that was enough.
That would confirm the cause.
This is the part I can't figure out from the code alone, which is why I'm asking.
I'll take care of the rest on my side.
Separately from that, I also want to give you a way out other than rebooting.
When a leftover backend is detected, I'd like to show a short explanation with a button, and pressing it would clean up what's left and bring the backend back up. Retry clearly isn't enough to get out of this situation.
Either way this needs some hardening, so I started sketching out more defensive measures.
I'll just leave the sketch here.
What I currently suspect
The plugin unpacks the webview and backend files into a per-plugin-version directory once, and serializes that step with a .lock file.
The code that waits for this lock has no time limit, so if something is still holding the lock, it waits forever.
There is a 30 second limit around it, so it ends up failing with Plugin resources not ready after 30 seconds.
On Windows, a file held open by a running process can't be deleted or overwritten.
So if a backend from the previous version is still alive, the new version can't put its files in place, and since closing the IDE doesn't end that process, only a reboot clears it.
I haven't reproduced it yet though, so this is still an unverified guess.
1. The actual fix
I want to stop waiting for the lock indefinitely.
If the lock can't be acquired within a set time, give up on it, use the already unpacked files if they are intact, and otherwise unpack somewhere else and run from there.
That way, startup isn't blocked even when something is holding the files.
2. A fallback — a way out without rebooting
Separately from the fix above, I think you need a way to get out of this yourself when it happens.
When a leftover backend from the previous version is detected, I want to show a short explanation of what happened along with a button.
Pressing it would clean up the leftover process and bring the backend back up.
There are two conditions I want to keep here.
First, whoever originally started the backend should be the one to start it again.
If the IDE started it, the IDE restarts it; if it was started from a terminal, the terminal does.
Otherwise nobody ends up responsible for that backend's lifetime.
Second, it only starts a new one after confirming the previous process is really gone.
Starting without that check would just collide with the old process still holding the port, and the same problem would repeat.
I'm planning to factor this clean-up-then-restart behavior out so it can be reused elsewhere.
A backend that stops responding isn't unique to this case.
Retry stays as it is, but in a situation like this one where it isn't enough, the button above would show up next to it.
@i-kulbitski I'm sorry. It took a long time to fix another critical bug (#359).
Since that bug needs to be deployed more quickly, we have postponed including this issue in the next release.
@yhk1038 It happened again when updating to 0.29.3. And at the time I did have a separate node.js process running in the terminal. But killing it did not help; only full reboot did.
Thank you so much for staying with this one until the end. 🙏
This is now fixed and released in v0.29.3.
You can see the change in PR #360.
There is one thing I would like to mention.
When you told me it happened again after updating to 0.29.3, the fix had actually not shipped yet.
The PR with the fix was merged a day after that comment.
So your report was not ignored. You were looking at a version that was not fixed yet.
The part where only a reboot helped is what made this bug drag on for so long.
Your note that closing the IDE was not enough, and only a full reboot worked, was the clue that finally pinned it down.
The latest version is now v0.29.4.
When you have time, please update the plugin.
Then please check if it works well for you now.
If it is fixed, a quick reply here would be great.
If the problem is still there, please let me know in a comment.
I will reopen it and raise the priority to fix it quickly.
Activity
i-kulbitski commentedon Aug 19, 2026
Seems to be fixed in 0.28.0
yhk1038 commentedon Aug 19, 2026
@i-kulbitski Thank you so much for not forgetting to let me know!
i-kulbitski commentedon Aug 25, 2026
This issue has reappeared after update to 0.29.1:

Rebooting Windows (not just CLion) fixes it, but I'd prefer not having to do it
yhk1038 commentedon Aug 26, 2026
@i-kulbitski
Yeah... it's kind of weird that it only gets fixed after a reboot.
If a reboot is what fixes it, something must survive closing the IDE.
My guess is that a backend process from the older version is still alive and holding on to the plugin files.
If that's what happens, the plugin should be able to survive it on its own, so that's the direction I'm going to fix.
There is just one thing I'd like to ask you to check.
Next time it happens, close CLion completely, open Task Manager, and see if a
node.exeis still there.If you keep your project inside WSL, it will show up as
wsl.exeinstead ofnode.exe.If it is there, end just that process and start CLion again without rebooting, and let me know whether that was enough.
That would confirm the cause.
This is the part I can't figure out from the code alone, which is why I'm asking.
I'll take care of the rest on my side.
Separately from that, I also want to give you a way out other than rebooting.
When a leftover backend is detected, I'd like to show a short explanation with a button, and pressing it would clean up what's left and bring the backend back up.
Retryclearly isn't enough to get out of this situation.I'm reopening this issue.
yhk1038 commentedon Aug 26, 2026
Either way this needs some hardening, so I started sketching out more defensive measures.
I'll just leave the sketch here.
What I currently suspect
The plugin unpacks the webview and backend files into a per-plugin-version directory once, and serializes that step with a
.lockfile.The code that waits for this lock has no time limit, so if something is still holding the lock, it waits forever.
There is a 30 second limit around it, so it ends up failing with
Plugin resources not readyafter 30 seconds.On Windows, a file held open by a running process can't be deleted or overwritten.
So if a backend from the previous version is still alive, the new version can't put its files in place, and since closing the IDE doesn't end that process, only a reboot clears it.
I haven't reproduced it yet though, so this is still an unverified guess.
1. The actual fix
I want to stop waiting for the lock indefinitely.
If the lock can't be acquired within a set time, give up on it, use the already unpacked files if they are intact, and otherwise unpack somewhere else and run from there.
That way, startup isn't blocked even when something is holding the files.
2. A fallback — a way out without rebooting
Separately from the fix above, I think you need a way to get out of this yourself when it happens.
When a leftover backend from the previous version is detected, I want to show a short explanation of what happened along with a button.
Pressing it would clean up the leftover process and bring the backend back up.
There are two conditions I want to keep here.
First, whoever originally started the backend should be the one to start it again.
If the IDE started it, the IDE restarts it; if it was started from a terminal, the terminal does.
Otherwise nobody ends up responsible for that backend's lifetime.
Second, it only starts a new one after confirming the previous process is really gone.
Starting without that check would just collide with the old process still holding the port, and the same problem would repeat.
I'm planning to factor this clean-up-then-restart behavior out so it can be reused elsewhere.
A backend that stops responding isn't unique to this case.
Retrystays as it is, but in a situation like this one where it isn't enough, the button above would show up next to it.yhk1038 commentedon Aug 27, 2026
@i-kulbitski I'm sorry. It took a long time to fix another critical bug (#359).
Since that bug needs to be deployed more quickly, we have postponed including this issue in the next release.
i-kulbitski commentedon Aug 27, 2026
@yhk1038 It happened again when updating to 0.29.3. And at the time I did have a separate node.js process running in the terminal. But killing it did not help; only full reboot did.
fix: stop a leftover backend from blocking startup until reboot
feat: end the backend when the host that spawned it goes away
yhk1038 commentedon Aug 31, 2026
Hi @i-kulbitski,
Thank you so much for staying with this one until the end. 🙏
This is now fixed and released in v0.29.3.
You can see the change in PR #360.
There is one thing I would like to mention.
When you told me it happened again after updating to 0.29.3, the fix had actually not shipped yet.
The PR with the fix was merged a day after that comment.
So your report was not ignored. You were looking at a version that was not fixed yet.
The part where only a reboot helped is what made this bug drag on for so long.
Your note that closing the IDE was not enough, and only a full reboot worked, was the clue that finally pinned it down.
The latest version is now v0.29.4.
When you have time, please update the plugin.
Then please check if it works well for you now.
If it is fixed, a quick reply here would be great.
If the problem is still there, please let me know in a comment.
I will reopen it and raise the priority to fix it quickly.
If it works well now, and if you have not already, a short review and rating would mean a lot to me: Leave a review on the JetBrains Marketplace
A GitHub star would also be very welcome and encouraging.
Thanks again for helping make the plugin better!
fix: stop the WSL2 backend from killing itself every ten seconds (#384)…