github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 00:51:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #15704] False "insufficient memory" error in LXC: Ollama appears to use MemFree instead of MemAvailable when loading models #72074

New Issue
Open
opened 2026-05-05 03:25:08 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-05 03:25:08 -05:00
Owner
Copy Link

Originally created by @avgex on GitHub (Apr 19, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15704

What is the issue?

Ollama refuses to load a model in an LXC container even though the container has enough reclaimable memory available.

Environment:

  • Proxmox LXC container
  • Debian guest
  • Container memory: 32 GiB
  • Container swap: 512 MiB
  • CPU inference only

Observed behavior:
When I try to load gemma4:26b, Ollama returns:

model requires more system memory (18.3 GiB) than is available (12.6 GiB)

However, inside the container:

  • MemFree is about 12 GiB
  • MemAvailable is about 31.9 GiB
  • almost no RAM is used by processes
  • about 21 GiB is file cache, mostly inactive_file, so it should be reclaimable by Linux

This suggests Ollama is using something close to MemFree + SwapFree for its pre-load memory check, while ignoring reclaimable page cache / MemAvailable.

Why I believe this is incorrect:

  • free -h shows:
    • free: ~12 GiB
    • available: ~31 GiB
  • /sys/fs/cgroup/memory.stat shows:
    • anon: ~35 MiB
    • file: ~21.2 GiB
    • inactive_file: ~21.1 GiB
  • ps aux --sort=-%mem shows no large memory-consuming processes

So the memory is not actually occupied by applications. It is mostly page cache.

Expected behavior:
Ollama should either:

  1. use MemAvailable instead of only MemFree, or
  2. account for reclaimable file cache, or
  3. provide an override / more aggressive loading option

Actual behavior:
Ollama refuses to load the model even though Linux should be able to reclaim enough cache to make room.

Notes:

  • Restarting only the ollama service does not help
  • Restarting the whole container may temporarily help because cache is cleared
  • This looks especially relevant for containerized environments such as LXC

Relevant log output

Apr 19 17:31:46 ollama ollama[199195]: time=2026-04-19T17:31:46.564+02:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="32.0 GiB" available="12.1 GiB"
Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.309+02:00 level=INFO source=sched.go:484 msg="system memory" total="32.0 GiB" free="12.1 GiB" free_swap="506.2 MiB"
Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=WARN source=server.go:1058 msg="model request too large for system" requested="18.3 GiB" available="12.6 GiB" total="32.0 GiB" free="12.1 GiB" swap="506.2 MiB"
Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="17.3 GiB"
Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="880.0 MiB"
Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="133.7 MiB"
Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=device.go:272 msg="total memory" size="18.3 GiB"
Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=sched.go:511 msg="Load failed" model=/usr/share/ollama/.ollama/models/blobs/sha256-7121486771cbfe218851513210c40b35dbdee93ab1ef43fe36283c883980f0df error="model requires more system memory (18.3 GiB) than is available (12.6 GiB)"
Apr 19 17:32:08 ollama ollama[199195]: [GIN] 2026/04/19 - 17:32:08 | 500 | 595.882098ms | 192.168.1.239 | POST "/api/chat"

Inside the container:

free -h
---------------
               total        used        free      shared  buff/cache   available
Mem:            32Gi        44Mi        12Gi        72Ki        19Gi        31Gi
Swap:          512Mi       5.8Mi       506Mi

/proc/meminfo
---------------
MemTotal:       33554432 kB
MemFree:        12726260 kB
MemAvailable:   33508662 kB
SwapTotal:        524288 kB
SwapFree:         518352 kB

/sys/fs/cgroup/memory.stat
---------------
anon 36884480
file 21266915328
inactive_file 21189849088
active_file 77045760

OS

Linux

GPU

No response

CPU

Intel

Ollama version

0.20.5

Originally created by @avgex on GitHub (Apr 19, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15704 ### What is the issue? Ollama refuses to load a model in an LXC container even though the container has enough reclaimable memory available. Environment: - Proxmox LXC container - Debian guest - Container memory: 32 GiB - Container swap: 512 MiB - CPU inference only Observed behavior: When I try to load `gemma4:26b`, Ollama returns: `model requires more system memory (18.3 GiB) than is available (12.6 GiB)` However, inside the container: - `MemFree` is about 12 GiB - `MemAvailable` is about 31.9 GiB - almost no RAM is used by processes - about 21 GiB is file cache, mostly `inactive_file`, so it should be reclaimable by Linux This suggests Ollama is using something close to `MemFree + SwapFree` for its pre-load memory check, while ignoring reclaimable page cache / `MemAvailable`. Why I believe this is incorrect: - `free -h` shows: - free: ~12 GiB - available: ~31 GiB - `/sys/fs/cgroup/memory.stat` shows: - anon: ~35 MiB - file: ~21.2 GiB - inactive_file: ~21.1 GiB - `ps aux --sort=-%mem` shows no large memory-consuming processes So the memory is not actually occupied by applications. It is mostly page cache. Expected behavior: Ollama should either: 1. use `MemAvailable` instead of only `MemFree`, or 2. account for reclaimable file cache, or 3. provide an override / more aggressive loading option Actual behavior: Ollama refuses to load the model even though Linux should be able to reclaim enough cache to make room. Notes: - Restarting only the `ollama` service does not help - Restarting the whole container may temporarily help because cache is cleared - This looks especially relevant for containerized environments such as LXC ### Relevant log output ```shell Apr 19 17:31:46 ollama ollama[199195]: time=2026-04-19T17:31:46.564+02:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="32.0 GiB" available="12.1 GiB" Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.309+02:00 level=INFO source=sched.go:484 msg="system memory" total="32.0 GiB" free="12.1 GiB" free_swap="506.2 MiB" Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=WARN source=server.go:1058 msg="model request too large for system" requested="18.3 GiB" available="12.6 GiB" total="32.0 GiB" free="12.1 GiB" swap="506.2 MiB" Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="17.3 GiB" Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="880.0 MiB" Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="133.7 MiB" Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=device.go:272 msg="total memory" size="18.3 GiB" Apr 19 17:32:08 ollama ollama[199195]: time=2026-04-19T17:32:08.568+02:00 level=INFO source=sched.go:511 msg="Load failed" model=/usr/share/ollama/.ollama/models/blobs/sha256-7121486771cbfe218851513210c40b35dbdee93ab1ef43fe36283c883980f0df error="model requires more system memory (18.3 GiB) than is available (12.6 GiB)" Apr 19 17:32:08 ollama ollama[199195]: [GIN] 2026/04/19 - 17:32:08 | 500 | 595.882098ms | 192.168.1.239 | POST "/api/chat" Inside the container: free -h --------------- total used free shared buff/cache available Mem: 32Gi 44Mi 12Gi 72Ki 19Gi 31Gi Swap: 512Mi 5.8Mi 506Mi /proc/meminfo --------------- MemTotal: 33554432 kB MemFree: 12726260 kB MemAvailable: 33508662 kB SwapTotal: 524288 kB SwapFree: 518352 kB /sys/fs/cgroup/memory.stat --------------- anon 36884480 file 21266915328 inactive_file 21189849088 active_file 77045760 ``` ### OS Linux ### GPU _No response_ ### CPU Intel ### Ollama version 0.20.5
GiteaMirror added the bug label 2026-05-05 03:25:08 -05:00
GiteaMirror commented 2026-05-05 03:25:10 -05:00
Author
Owner
Copy Link

@two56bytes commented on GitHub (Apr 21, 2026):

+1

<!-- gh-comment-id:4288248898 --> @two56bytes commented on GitHub (Apr 21, 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 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#72074
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?