github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 16:40:08 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #15419] Ollama Cloud: Frequent 503 errors making cloud models unreliable #71918

New Issue
Open
opened 2026-05-05 02:56:47 -05:00 by GiteaMirror · 18 comments
GiteaMirror commented 2026-05-05 02:56:47 -05:00
Owner
Copy Link

Originally created by @oggixx on GitHub (Apr 8, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15419

Hey Ollama team,

I've been running autonomous agents using Ollama Cloud models and running into frequent 503 Service Unavailable errors. Thought I'd report this since it's making the cloud models pretty unreliable for production use.

What's happening

Requests to cloud models randomly fail with:

This happens with:

  • glm-5:cloud
  • glm-5.1:cloud
  • minimax-m2.7:cloud
  • kimi-k2.5:cloud

How often

Multiple times per hour. Sometimes a single request works, the next one fails with 503, then it works again. It's intermittent but frequent enough to break agent workflows.

Impact

  • Agent requests timeout mid-conversation
  • Cron jobs fail unexpectedly
  • Users experience broken interactions in Telegram/Discord bots

What I've tried

  • Retry logic with exponential backoff
  • Falling back to different models
  • None of it really helps when the 503s are this frequent

Setup

  • Ollama running in Docker
  • Using the model suffix to hit ollama.com:443
  • Fairly standard setup otherwise

Would be helpful

  1. Some visibility into when/why 503s happen (rate limiting vs capacity issues?)
  2. A retry-after header so we can back off properly
  3. Maybe a status page for Ollama Cloud uptime?

Thanks for the great work on Ollama! Happy to provide more details if needed.

Originally created by @oggixx on GitHub (Apr 8, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15419 Hey Ollama team, I've been running autonomous agents using Ollama Cloud models and running into frequent 503 Service Unavailable errors. Thought I'd report this since it's making the cloud models pretty unreliable for production use. ## What's happening Requests to cloud models randomly fail with: This happens with: - glm-5:cloud - glm-5.1:cloud - minimax-m2.7:cloud - kimi-k2.5:cloud ## How often Multiple times per hour. Sometimes a single request works, the next one fails with 503, then it works again. It's intermittent but frequent enough to break agent workflows. ## Impact - Agent requests timeout mid-conversation - Cron jobs fail unexpectedly - Users experience broken interactions in Telegram/Discord bots ## What I've tried - Retry logic with exponential backoff - Falling back to different models - None of it really helps when the 503s are this frequent ## Setup - Ollama running in Docker - Using the model suffix to hit ollama.com:443 - Fairly standard setup otherwise ## Would be helpful 1. Some visibility into when/why 503s happen (rate limiting vs capacity issues?) 2. A retry-after header so we can back off properly 3. Maybe a status page for Ollama Cloud uptime? Thanks for the great work on Ollama! Happy to provide more details if needed.
GiteaMirror added the cloud label 2026-05-05 02:56:47 -05:00
GiteaMirror commented 2026-05-05 02:56:49 -05:00
Author
Owner
Copy Link

@patelhiren commented on GitHub (Apr 8, 2026):

Experienced the same this morning. Frequent 503 errors but not consistent. A few requests fail with 503 then get one successful one and so on.

<!-- gh-comment-id:4207079804 --> @patelhiren commented on GitHub (Apr 8, 2026): Experienced the same this morning. Frequent 503 errors but not consistent. A few requests fail with 503 then get one successful one and so on.
GiteaMirror commented 2026-05-05 02:56:50 -05:00
Author
Owner
Copy Link

@MetaSep commented on GitHub (Apr 8, 2026):

Exact same issue, glm5.1 is unusable with open code and openclaw, consistent 503 errors.

<!-- gh-comment-id:4207374218 --> @MetaSep commented on GitHub (Apr 8, 2026): Exact same issue, glm5.1 is unusable with open code and openclaw, consistent 503 errors.
GiteaMirror commented 2026-05-05 02:56:52 -05:00
Author
Owner
Copy Link

@srflaherty commented on GitHub (Apr 8, 2026):

Same to report here on Qwen3.5 397B Cloud, opencode, and also noticed the Ollama Desktop app is struggling on my system to run the server, just hangs up after last update showing account trying to load but just sits there in that state. 503 down to even terminal calls to Ollama on my setup

<!-- gh-comment-id:4208294941 --> @srflaherty commented on GitHub (Apr 8, 2026): Same to report here on Qwen3.5 397B Cloud, opencode, and also noticed the Ollama Desktop app is struggling on my system to run the server, just hangs up after last update showing account trying to load but just sits there in that state. 503 down to even terminal calls to Ollama on my setup
GiteaMirror commented 2026-05-05 02:56:52 -05:00
Author
Owner
Copy Link

@backamblock commented on GitHub (Apr 8, 2026):

+1

<!-- gh-comment-id:4208874619 --> @backamblock commented on GitHub (Apr 8, 2026): +1
GiteaMirror commented 2026-05-05 02:56:55 -05:00
Author
Owner
Copy Link

@boeserwolf commented on GitHub (Apr 8, 2026):

+1

<!-- gh-comment-id:4210006064 --> @boeserwolf commented on GitHub (Apr 8, 2026): +1
GiteaMirror commented 2026-05-05 02:56:56 -05:00
Author
Owner
Copy Link

@Hevoon commented on GitHub (Apr 9, 2026):

#15290 same issue.

<!-- gh-comment-id:4211300852 --> @Hevoon commented on GitHub (Apr 9, 2026): #15290 same issue.
GiteaMirror commented 2026-05-05 02:56:57 -05:00
Author
Owner
Copy Link

@saltwater-tensor commented on GitHub (Apr 9, 2026):

+1
While using with vscode copilot agent mode (glm5.1) it is hitting 503 and 502 intermittently.

<!-- gh-comment-id:4213211189 --> @saltwater-tensor commented on GitHub (Apr 9, 2026): +1 While using with vscode copilot agent mode (glm5.1) it is hitting 503 and 502 intermittently.
GiteaMirror commented 2026-05-05 02:56:59 -05:00
Author
Owner
Copy Link

@imaboku commented on GitHub (Apr 10, 2026):

Same issues this week with Kimi2.5 and GLM5.1 w Claude Code. It would appear restarting restarting Ollama on my local machine resolves the issue for some time but after continuing a long session it will happen again. Have not run ollama server in debug mode yet to see if I can see more details on the error. Error that starts to show when issues arise: level=WARN source=cloud_proxy.go:257 msg="cloud proxy response copy failed" path=/v1/messages upstream_path=/v1/messages status=200 request_context_canceled=false request_context_err=<nil> error="unexpected EOF"

<!-- gh-comment-id:4220308932 --> @imaboku commented on GitHub (Apr 10, 2026): Same issues this week with Kimi2.5 and GLM5.1 w Claude Code. It would appear restarting restarting Ollama on my local machine resolves the issue for some time but after continuing a long session it will happen again. Have not run ollama server in debug mode yet to see if I can see more details on the error. Error that starts to show when issues arise: ` level=WARN source=cloud_proxy.go:257 msg="cloud proxy response copy failed" path=/v1/messages upstream_path=/v1/messages status=200 request_context_canceled=false request_context_err=<nil> error="unexpected EOF"`
GiteaMirror commented 2026-05-05 02:57:00 -05:00
Author
Owner
Copy Link

@Henkypenky1 commented on GitHub (Apr 10, 2026):

+1 , set up fallback models glm 5.1>glm 5> minimax m2.5 and landing frequently on minimax.

<!-- gh-comment-id:4226085704 --> @Henkypenky1 commented on GitHub (Apr 10, 2026): +1 , set up fallback models glm 5.1>glm 5> minimax m2.5 and landing frequently on minimax.
GiteaMirror commented 2026-05-05 02:57:04 -05:00
Author
Owner
Copy Link

@orrinwitt commented on GitHub (Apr 12, 2026):

Also running into this a lot lately on minimax-m2.7 and glm-5.1 using nanobot

<!-- gh-comment-id:4232615125 --> @orrinwitt commented on GitHub (Apr 12, 2026): Also running into this a lot lately on minimax-m2.7 and glm-5.1 using nanobot
GiteaMirror commented 2026-05-05 02:57:06 -05:00
Author
Owner
Copy Link

@brunobarbarioli commented on GitHub (Apr 13, 2026):

same using glm-5.1 . It has become unusable at this point

<!-- gh-comment-id:4236532664 --> @brunobarbarioli commented on GitHub (Apr 13, 2026): same using glm-5.1 . It has become unusable at this point
GiteaMirror commented 2026-05-05 02:57:08 -05:00
Author
Owner
Copy Link

@backamblock commented on GitHub (Apr 13, 2026):

puh.. now the other days it was slow, but today its just constantly dropping the connections.. how are we supposed to work with this?
funny this just started after i bought 200$ yearly sub for a colleague and me lol. -.-

is there something we can do on our end as users to make it better? other than not using it to reduce server load.

my failing models at the moment are qwen3.5, minimax2.7, glm5.1

<!-- gh-comment-id:4237161481 --> @backamblock commented on GitHub (Apr 13, 2026): puh.. now the other days it was slow, but today its just constantly dropping the connections.. how are we supposed to work with this? funny this just started after i bought 200$ yearly sub for a colleague and me lol. -.- is there something we can do on our end as users to make it better? other than not using it to reduce server load. my failing models at the moment are qwen3.5, minimax2.7, glm5.1
GiteaMirror commented 2026-05-05 02:57:10 -05:00
Author
Owner
Copy Link

@itisaevalex commented on GitHub (Apr 13, 2026):

+1

<!-- gh-comment-id:4238047284 --> @itisaevalex commented on GitHub (Apr 13, 2026): +1
GiteaMirror commented 2026-05-05 02:57:10 -05:00
Author
Owner
Copy Link

@danielithomas commented on GitHub (Apr 16, 2026):

Watching - Service status for cloud models would be useful.

<!-- gh-comment-id:4257704002 --> @danielithomas commented on GitHub (Apr 16, 2026): Watching - Service status for cloud models would be useful.
GiteaMirror commented 2026-05-05 02:57:11 -05:00
Author
Owner
Copy Link

@PureBlissAK commented on GitHub (Apr 18, 2026):

🤖 Automated Triage & Analysis Report

Issue: #15419
Analyzed: 2026-04-18T18:21:34.290458

Analysis

  • Type: unknown
  • Severity: medium
  • Components: unknown

Implementation Plan

  • Effort: medium
  • Steps:

This issue has been triaged and marked for implementation.

<!-- gh-comment-id:4274308560 --> @PureBlissAK commented on GitHub (Apr 18, 2026): <!-- ollama-issue-orchestrator:v1 issue:15419 --> ## 🤖 Automated Triage & Analysis Report **Issue**: #15419 **Analyzed**: 2026-04-18T18:21:34.290458 ### Analysis - **Type**: unknown - **Severity**: medium - **Components**: unknown ### Implementation Plan - **Effort**: medium - **Steps**: *This issue has been triaged and marked for implementation.*
GiteaMirror commented 2026-05-05 02:57:11 -05:00
Author
Owner
Copy Link

@Dawud18854 commented on GitHub (Apr 21, 2026):

+1

<!-- gh-comment-id:4289873611 --> @Dawud18854 commented on GitHub (Apr 21, 2026): +1
GiteaMirror commented 2026-05-05 02:57:12 -05:00
Author
Owner
Copy Link

@Ner-Kun commented on GitHub (Apr 29, 2026):

+1

<!-- gh-comment-id:4346385794 --> @Ner-Kun commented on GitHub (Apr 29, 2026): +1
GiteaMirror commented 2026-05-05 02:57:12 -05:00
Author
Owner
Copy Link

@dougdyson commented on GitHub (May 3, 2026):

+1

<!-- gh-comment-id:4366880563 --> @dougdyson commented on GitHub (May 3, 2026): +1
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 cloud
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#71918
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?