github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 08:02:14 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #8214] /api/chat and /api/generate endpoints are timing out #67299

New Issue
Closed
opened 2026-05-04 09:53:59 -05:00 by GiteaMirror · 13 comments
GiteaMirror commented 2026-05-04 09:53:59 -05:00
Owner
Copy Link

Originally created by @wkevin on GitHub (Dec 23, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/8214

What is the issue?

Currently, the /api/tags and /api/ps endpoints are functioning properly, but the /api/chat and /api/generate endpoints are experiencing timeouts.

Docker Image Version: 0.5.4

Upon checking the server information, I noticed that the VRAM usage and processes reported by nvidia-smi do not match those shown by the ollama ps command.

image

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.5.4

Originally created by @wkevin on GitHub (Dec 23, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/8214 ### What is the issue? Currently, the `/api/tags` and `/api/ps` endpoints are functioning properly, but the `/api/chat` and `/api/generate` endpoints are experiencing timeouts. Docker Image Version: 0.5.4 Upon checking the server information, I noticed that the VRAM usage and processes reported by `nvidia-smi` do not match those shown by the `ollama ps` command. ![image](https://github.com/user-attachments/assets/17ac4c3f-097c-410c-9c62-b388c0ede2c1) ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.5.4
GiteaMirror added the bug label 2026-05-04 09:53:59 -05:00
GiteaMirror commented 2026-05-04 09:54:02 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Dec 23, 2024):

Server logs will aid in debugging.

<!-- gh-comment-id:2558781810 --> @rick-github commented on GitHub (Dec 23, 2024): [Server logs](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.md#how-to-troubleshoot-issues) will aid in debugging.
GiteaMirror commented 2026-05-04 09:54:02 -05:00
Author
Owner
Copy Link

@wkevin commented on GitHub (Dec 23, 2024):

Server logs will aid in debugging.

I have rolled back to the previous version docker image, so docker logs has gone.
Now, switching to 0.5.4 reproduce this issue, Please wait. Thanks.

<!-- gh-comment-id:2558789981 --> @wkevin commented on GitHub (Dec 23, 2024): > [Server logs](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.md#how-to-troubleshoot-issues) will aid in debugging. I have rolled back to the previous version docker image, so docker logs has gone. Now, switching to 0.5.4 reproduce this issue, Please wait. Thanks.
GiteaMirror commented 2026-05-04 09:54:02 -05:00
Author
Owner
Copy Link

@a1ananth commented on GitHub (Dec 23, 2024):

I am having this same issue with v 0.5.4. It works fine if I downgrade, but due to auto-update, it crashes again.

My logs are:

time=2024-12-23T13:31:57.296+05:30 level=INFO source=sched.go:714 msg="new model will fit in available VRAM in single GPU, loading" model=E:\AI\Ollama\models\blobs\sha256-60e05f2100071479f596b964f89f510f057ce397ea22f2833a0cfe029bfc2463 gpu=GPU-345e1da7-095d-c43e-50f9-cb9f21b7d201 parallel=4 available=11379396608 required="5.6 GiB"
time=2024-12-23T13:31:57.314+05:30 level=INFO source=server.go:104 msg="system memory" total="63.9 GiB" free="41.7 GiB" free_swap="49.2 GiB"
time=2024-12-23T13:31:57.314+05:30 level=INFO source=memory.go:356 msg="offload to cuda" layers.requested=-1 layers.model=29 layers.offload=29 layers.split="" memory.available="[10.6 GiB]" memory.gpu_overhead="0 B" memory.required.full="5.6 GiB" memory.required.partial="5.6 GiB" memory.required.kv="448.0 MiB" memory.required.allocations="[5.6 GiB]" memory.weights.total="4.1 GiB" memory.weights.repeating="3.7 GiB" memory.weights.nonrepeating="426.4 MiB" memory.graph.full="478.0 MiB" memory.graph.partial="730.4 MiB"
time=2024-12-23T13:31:57.315+05:30 level=INFO source=server.go:160 msg="user override" OLLAMA_LLM_LIBRARY=cuda_v12 path=D:\Programs\AI\ollama\lib\ollama\runners\cuda_v12\ollama_llama_server.exe
time=2024-12-23T13:31:57.317+05:30 level=INFO source=server.go:376 msg="starting llama server" cmd="D:\\Programs\\AI\\ollama\\lib\\ollama\\runners\\cuda_v12\\ollama_llama_server.exe runner --model E:\\AI\\Ollama\\models\\blobs\\sha256-60e05f2100071479f596b964f89f510f057ce397ea22f2833a0cfe029bfc2463 --ctx-size 8192 --batch-size 512 --n-gpu-layers 29 --threads 8 --no-mmap --parallel 4 --port 59599"
time=2024-12-23T13:31:57.320+05:30 level=INFO source=sched.go:449 msg="loaded runners" count=1
time=2024-12-23T13:31:57.320+05:30 level=INFO source=server.go:555 msg="waiting for llama runner to start responding"
time=2024-12-23T13:31:57.320+05:30 level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server error"
time=2024-12-23T13:31:57.389+05:30 level=INFO source=runner.go:941 msg="starting go runner"
time=2024-12-23T13:31:57.389+05:30 level=INFO source=runner.go:942 msg=system info="AVX = 1 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 0 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 0 | FP16_VA = 0 | RISCV_VECT = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 | cgo(clang)" threads=8
time=2024-12-23T13:31:57.390+05:30 level=INFO source=.:0 msg="Server listening on 127.0.0.1:8080"
gguf_init_from_file: failed to open '': 'Invalid argument'
llama_model_load: error loading model: llama_model_loader: failed to load model from 

llama_load_model_from_file: failed to load model
panic: unable to load model: 

goroutine 6 [running]:
main.(*Server).loadModel(0xc0000a8120, {0x0, 0x0, 0x1, 0x0, {0x0, 0x0, 0x0}, 0xc0000241c0, 0x0}, ...)
	github.com/ollama/ollama/llama/runner/runner.go:863 +0x3ad
created by main.main in goroutine 1
	github.com/ollama/ollama/llama/runner/runner.go:975 +0xc6c
time=2024-12-23T13:31:57.571+05:30 level=ERROR source=sched.go:455 msg="error loading llama server" error="llama runner process has terminated: error loading model: llama_model_loader: failed to load model from"
[GIN] 2024/12/23 - 13:31:57 | 500 |     335.213ms |  192.168.29.160 | POST     "/v1/chat/completions"

This repeats 3 times on each completion request, possibly due to retry logic.

<!-- gh-comment-id:2559115338 --> @a1ananth commented on GitHub (Dec 23, 2024): I am having this same issue with v 0.5.4. It works fine if I downgrade, but due to auto-update, it crashes again. My logs are: ``` time=2024-12-23T13:31:57.296+05:30 level=INFO source=sched.go:714 msg="new model will fit in available VRAM in single GPU, loading" model=E:\AI\Ollama\models\blobs\sha256-60e05f2100071479f596b964f89f510f057ce397ea22f2833a0cfe029bfc2463 gpu=GPU-345e1da7-095d-c43e-50f9-cb9f21b7d201 parallel=4 available=11379396608 required="5.6 GiB" time=2024-12-23T13:31:57.314+05:30 level=INFO source=server.go:104 msg="system memory" total="63.9 GiB" free="41.7 GiB" free_swap="49.2 GiB" time=2024-12-23T13:31:57.314+05:30 level=INFO source=memory.go:356 msg="offload to cuda" layers.requested=-1 layers.model=29 layers.offload=29 layers.split="" memory.available="[10.6 GiB]" memory.gpu_overhead="0 B" memory.required.full="5.6 GiB" memory.required.partial="5.6 GiB" memory.required.kv="448.0 MiB" memory.required.allocations="[5.6 GiB]" memory.weights.total="4.1 GiB" memory.weights.repeating="3.7 GiB" memory.weights.nonrepeating="426.4 MiB" memory.graph.full="478.0 MiB" memory.graph.partial="730.4 MiB" time=2024-12-23T13:31:57.315+05:30 level=INFO source=server.go:160 msg="user override" OLLAMA_LLM_LIBRARY=cuda_v12 path=D:\Programs\AI\ollama\lib\ollama\runners\cuda_v12\ollama_llama_server.exe time=2024-12-23T13:31:57.317+05:30 level=INFO source=server.go:376 msg="starting llama server" cmd="D:\\Programs\\AI\\ollama\\lib\\ollama\\runners\\cuda_v12\\ollama_llama_server.exe runner --model E:\\AI\\Ollama\\models\\blobs\\sha256-60e05f2100071479f596b964f89f510f057ce397ea22f2833a0cfe029bfc2463 --ctx-size 8192 --batch-size 512 --n-gpu-layers 29 --threads 8 --no-mmap --parallel 4 --port 59599" time=2024-12-23T13:31:57.320+05:30 level=INFO source=sched.go:449 msg="loaded runners" count=1 time=2024-12-23T13:31:57.320+05:30 level=INFO source=server.go:555 msg="waiting for llama runner to start responding" time=2024-12-23T13:31:57.320+05:30 level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server error" time=2024-12-23T13:31:57.389+05:30 level=INFO source=runner.go:941 msg="starting go runner" time=2024-12-23T13:31:57.389+05:30 level=INFO source=runner.go:942 msg=system info="AVX = 1 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 0 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 0 | FP16_VA = 0 | RISCV_VECT = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 | cgo(clang)" threads=8 time=2024-12-23T13:31:57.390+05:30 level=INFO source=.:0 msg="Server listening on 127.0.0.1:8080" gguf_init_from_file: failed to open '': 'Invalid argument' llama_model_load: error loading model: llama_model_loader: failed to load model from llama_load_model_from_file: failed to load model panic: unable to load model: goroutine 6 [running]: main.(*Server).loadModel(0xc0000a8120, {0x0, 0x0, 0x1, 0x0, {0x0, 0x0, 0x0}, 0xc0000241c0, 0x0}, ...) github.com/ollama/ollama/llama/runner/runner.go:863 +0x3ad created by main.main in goroutine 1 github.com/ollama/ollama/llama/runner/runner.go:975 +0xc6c time=2024-12-23T13:31:57.571+05:30 level=ERROR source=sched.go:455 msg="error loading llama server" error="llama runner process has terminated: error loading model: llama_model_loader: failed to load model from" [GIN] 2024/12/23 - 13:31:57 | 500 | 335.213ms | 192.168.29.160 | POST "/v1/chat/completions" ``` This repeats 3 times on each completion request, possibly due to retry logic.
GiteaMirror commented 2026-05-04 09:54:03 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Dec 23, 2024):

It's not the same issue, your runners are crashing rather than going slow. Create a new issue and include full server logs.

<!-- gh-comment-id:2559136088 --> @rick-github commented on GitHub (Dec 23, 2024): It's not the same issue, your runners are crashing rather than going slow. Create a new issue and include full server logs.
GiteaMirror commented 2026-05-04 09:54:03 -05:00
Author
Owner
Copy Link

@jessegross commented on GitHub (Dec 23, 2024):

@wkevin What was the previous version that you rolled back to that didn't have the issue?

<!-- gh-comment-id:2560050514 --> @jessegross commented on GitHub (Dec 23, 2024): @wkevin What was the previous version that you rolled back to that didn't have the issue?
GiteaMirror commented 2026-05-04 09:54:04 -05:00
Author
Owner
Copy Link

@wkevin commented on GitHub (Dec 25, 2024):

@wkevin What was the previous version that you rolled back to that didn't have the issue?

0.3.10

I have run 0.5.4 again for two days, and timeout hasn't reoccurred。

<!-- gh-comment-id:2561926796 --> @wkevin commented on GitHub (Dec 25, 2024): > @wkevin What was the previous version that you rolled back to that didn't have the issue? 0.3.10 I have run 0.5.4 again for two days, and timeout hasn't reoccurred。
GiteaMirror commented 2026-05-04 09:54:05 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Dec 26, 2024):

I speculate that the problem @wkevin had was a GPU disconnect (https://github.com/ollama/ollama/issues/6928) or some other Nvidia error. The runners stay loaded but because they can no longer use the GPU, they fallback to doing inference on the CPU. If the client has a deadline that assumes a GPU is doing inference, then it may be too short for the case where the CPU is being used. What deadlines have you set on the API calls?

<!-- gh-comment-id:2562032298 --> @rick-github commented on GitHub (Dec 26, 2024): I speculate that the problem @wkevin had was a GPU disconnect (https://github.com/ollama/ollama/issues/6928) or some other Nvidia error. The runners stay loaded but because they can no longer use the GPU, they fallback to doing inference on the CPU. If the client has a deadline that assumes a GPU is doing inference, then it may be too short for the case where the CPU is being used. What deadlines have you set on the API calls?
GiteaMirror commented 2026-05-04 09:54:06 -05:00
Author
Owner
Copy Link

@wkevin commented on GitHub (Dec 26, 2024):

I speculate that the problem @wkevin had was a GPU disconnect (#6928) or some other Nvidia error. The runners stay loaded but because they can no longer use the GPU, they fallback to doing inference on the CPU. If the client has a deadline that assumes a GPU is doing inference, then it may be too short for the case where the CPU is being used. What deadlines have you set on the API calls?

Thanks. But process is 100% GPU in ollama ps output.

I use VSCode + Continue, no specialized configuration of deadlines.

<!-- gh-comment-id:2562104567 --> @wkevin commented on GitHub (Dec 26, 2024): > I speculate that the problem @wkevin had was a GPU disconnect (#6928) or some other Nvidia error. The runners stay loaded but because they can no longer use the GPU, they fallback to doing inference on the CPU. If the client has a deadline that assumes a GPU is doing inference, then it may be too short for the case where the CPU is being used. What deadlines have you set on the API calls? Thanks. But process is 100% GPU in `ollama ps` output. I use VSCode + Continue, no specialized configuration of deadlines.
GiteaMirror commented 2026-05-04 09:54:07 -05:00
Author
Owner
Copy Link

@wkevin commented on GitHub (Dec 26, 2024):

The good news is that qwen2.5-coder:14b on the 0.5.4 is much faster to generate than 0.3.10. Many thanks to the ollama team!

<!-- gh-comment-id:2562105890 --> @wkevin commented on GitHub (Dec 26, 2024): The good news is that qwen2.5-coder:14b on the 0.5.4 is much faster to generate than 0.3.10. Many thanks to the ollama team!
GiteaMirror commented 2026-05-04 09:54:07 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Dec 26, 2024):

Thanks. But process is 100% GPU in ollama ps output.

Yes, the models were loaded when the GPU was accessible but then it was disconnected. Server logs would show what happened.

I use VSCode + Continue, no specialized configuration of deadlines.

What deadline does VSCode + Continue set?

<!-- gh-comment-id:2562108951 --> @rick-github commented on GitHub (Dec 26, 2024): > Thanks. But process is 100% GPU in ollama ps output. Yes, the models were loaded when the GPU was accessible but then it was disconnected. Server logs would show what happened. > I use VSCode + Continue, no specialized configuration of deadlines. What deadline does VSCode + Continue set?
GiteaMirror commented 2026-05-04 09:54:07 -05:00
Author
Owner
Copy Link

@wkevin commented on GitHub (Dec 26, 2024):

What deadline does VSCode + Continue set?

I don't know how to config API call's deadline in VSCode + Continue.

<!-- gh-comment-id:2562232689 --> @wkevin commented on GitHub (Dec 26, 2024): > What deadline does VSCode + Continue set? I don't know how to config API call's deadline in VSCode + Continue.
GiteaMirror commented 2026-05-04 09:54:08 -05:00
Author
Owner
Copy Link

@wkevin commented on GitHub (Dec 30, 2024):

I have run 0.5.4 for one week, timeout hasn't reoccurred。
Maybe this issue should close.

<!-- gh-comment-id:2564999692 --> @wkevin commented on GitHub (Dec 30, 2024): I have run 0.5.4 for one week, timeout hasn't reoccurred。 Maybe this issue should close.
GiteaMirror commented 2026-05-04 09:54:09 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Dec 30, 2024):

If the problem repeats, get the server logs before rolling back and re-open this issue.

<!-- gh-comment-id:2565135439 --> @rick-github commented on GitHub (Dec 30, 2024): If the problem repeats, get the server logs before rolling back and re-open this issue.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#67299
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?