-
-
Notifications
You must be signed in to change notification settings - Fork 1.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Killing idle backend failure. #1760
Comments
good catch, I can't confirm that yet - but look likes a mutex issue |
seem i have the same issue.. |
mudler
added a commit
that referenced
this issue
Mar 23, 2024
should be fixed in #1882 |
mudler
added a commit
that referenced
this issue
Mar 23, 2024
@TwinFinz can you try again from master when the builds are finished? Feel free to re-open if problem persists |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
LocalAI version:
sha-ff88c39-cublas-cuda12-ffmpeg (docker image)
Environment, CPU architecture, OS, and Version:
(Standard docker image)
Linux LocalAi 6.5.11-8-pve #1 SMP PREEMPT_DYNAMIC PMX 6.5.11-8 (2024-01-30T12:27Z) x86_64 x86_64 x86_64 GNU/Linux
Describe the bug
Killing idle llama backend from watchdog causes invalid pointer crash.
To Reproduce
Set enviorment variables:
Load multiple (2+) models with llama in succession then let it idle allowing watchdog to kill the backend.
Expected behavior
Backend should shutdown gracefully.
Logs
11:41PM WRN [WatchDog] Address 127.0.0.1:35825 is idle for too long, killing it
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x30 pc=0xb95344]
goroutine 39 [running]:
github.com/mudler/go-processmanager.(*Process).path(...)
/root/go/pkg/mod/github.com/mudler/go-processmanager@v0.0.0-20230818213616-f204007f963c/process.go:32
github.com/mudler/go-processmanager.(*Process).readPID(0x0)
/root/go/pkg/mod/github.com/mudler/go-processmanager@v0.0.0-20230818213616-f204007f963c/process.go:52 +0x24
github.com/mudler/go-processmanager.(*Process).Stop(0x0)
/root/go/pkg/mod/github.com/mudler/go-processmanager@v0.0.0-20230818213616-f204007f963c/process.go:150 +0x1c
github.com/go-skynet/LocalAI/pkg/model.(*ModelLoader).deleteProcess(0xc0004dee00, {0xc000a3a200, 0x20})
/build/pkg/model/process.go:32 +0x45
github.com/go-skynet/LocalAI/pkg/model.(*ModelLoader).StopModel(0xd1f720?, {0xc000a3a200, 0x20})
/build/pkg/model/loader.go:160 +0x111
github.com/go-skynet/LocalAI/pkg/model.(*WatchDog).checkIdle(0xc0005d4240)
/build/pkg/model/watchdog.go:115 +0x2c2
github.com/go-skynet/LocalAI/pkg/model.(*WatchDog).Run(0xc0005d4240)
/build/pkg/model/watchdog.go:99 +0xdd
created by github.com/go-skynet/LocalAI/core/http.Startup in goroutine 1
/build/core/http/api.go:100 +0x8fe
Additional context
This only tends to happen when multiple models are called, when it is only a single model loaded regardless of requests sent the crash does not occur.
The text was updated successfully, but these errors were encountered: