github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #11636] Retries with same text hang continually without modification #33449

New Issue
Closed
opened 2026-04-22 16:08:06 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-22 16:08:06 -05:00
Owner
Copy Link

Originally created by @leonletto on GitHub (Aug 1, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11636

What is the issue?

I'm not sure what category of bug this is but I have been fighting it for a long time ( months ) and finally worked around it today.

Steps to recreate:
1 You send an /api/generate POST call and the model hangs during inference causing a timeout ( for whatever reason ) .
2 you send the exact same call retrying, but inference hangs again
3 repeat ...

I think its something about the caching in the internal state causing it to repeat and hang in the same place but thats just me thinking out loud.

In the past I was restarting my whole chain and sometimes ollama to fix this but I got it working today.

If you send an /api/generate POST call and the model hangs during inference causing a timeout, before sending the retry you need to modify the original in a small way so that Ollama and/or the model think its a brand new call. Then it will recompute everything and you are good to go.

I guess this is a bug report and a workaround in one :)

Leon

Relevant log output

OS

No response

GPU

No response

CPU

No response

Ollama version

No response

Originally created by @leonletto on GitHub (Aug 1, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11636 ### What is the issue? I'm not sure what category of bug this is but I have been fighting it for a long time ( months ) and finally worked around it today. Steps to recreate: 1 You send an /api/generate POST call and the model hangs during inference causing a timeout ( for whatever reason ) . 2 you send the exact same call retrying, but inference hangs again 3 repeat ... I think its something about the caching in the internal state causing it to repeat and hang in the same place but thats just me thinking out loud. In the past I was restarting my whole chain and sometimes ollama to fix this but I got it working today. If you send an /api/generate POST call and the model hangs during inference causing a timeout, before sending the retry you need to modify the original in a small way so that Ollama and/or the model think its a brand new call. Then it will recompute everything and you are good to go. I guess this is a bug report and a workaround in one :) Leon ### Relevant log output ```shell ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version _No response_
GiteaMirror added the bug label 2026-04-22 16:08:06 -05:00
GiteaMirror commented 2026-04-22 16:08:08 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 2, 2025):

Example text?

<!-- gh-comment-id:3146048606 --> @rick-github commented on GitHub (Aug 2, 2025): Example text?
GiteaMirror commented 2026-04-22 16:08:08 -05:00
Author
Owner
Copy Link

@leonletto commented on GitHub (Aug 2, 2025):

I did a whole bunch of testing today to recreate the bug and I was able to do that. Ir turns out that if you reuse sessions over many calls ( as you would think would be efficient ), and because I am sending large prompts, I am getting HTTP session state corruption.

So my solution solved it even though I did not know why before. Anyway - looks like not a bug in Ollama but in plain old HTTP.

The end result - if you are in production and sending many prompts, don't reuse sessions. The time to setup a new http session is so small compared to inference time its not worth the chance of errors.

<!-- gh-comment-id:3146846359 --> @leonletto commented on GitHub (Aug 2, 2025): I did a whole bunch of testing today to recreate the bug and I was able to do that. Ir turns out that if you reuse sessions over many calls ( as you would think would be efficient ), and because I am sending large prompts, I am getting HTTP session state corruption. So my solution solved it even though I did not know why before. Anyway - looks like not a bug in Ollama but in plain old HTTP. The end result - if you are in production and sending many prompts, don't reuse sessions. The time to setup a new http session is so small compared to inference time its not worth the chance of errors.
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#33449
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?