Skip to content

fly.toml said the agent scales to zero; it has not since the LiteLLM move - #304

Merged
Sbussiso merged 1 commit into
masterfrom
docs/agent-is-not-scale-to-zero
Sep 14, 2026
Merged

Sbussiso merged 1 commit into
masterfrom
docs/agent-is-not-scale-to-zero

Conversation

@Sbussiso

Copy link
Copy Markdown
Contributor

The [processes] header described the agent as "scales to zero". The service
block sixty lines below says otherwise, and so does the machine:

fly.toml header:        "No volume, scales to zero"
agent service block:    auto_stop_machines = 'off'
                        min_machines_running = 1
live machine state:     started

Same failure as the cost comment fixed on 2026-09-12 — in the same file,
which I clearly didn't read closely enough while fixing that one. A file that
contradicts itself is worse than one that says nothing, because the reader
can't tell which half is current, and here the two halves are far enough apart
that you'd believe the first one and stop reading.

The reasoning for keeping it warm is sound and already argued beside the
service block: Fly's proxy gives an auto-started machine ~8s to bind, and this
process needed ~10s after LiteLLM landed. Scaling to zero is the shape this
worker wants; it isn't the shape it has.

Swept the rest of the tree for related stale claims — docs/README.md's "two
services scale to zero and two don't" is correct (License and Sync do; app and
agent don't).

flyctl config validate passes.

🤖 Generated with Claude Code

…move

The [processes] header described the agent as "scales to zero", while the
service block sixty lines below sets auto_stop_machines = "off" with
min_machines_running = 1, and the machine has been `started` continuously.

Same failure as the cost comment fixed on 2026-09-12 — in the same file,
which I clearly did not read closely enough while fixing that one. A file
that contradicts itself is worse than one that says nothing, because the
reader cannot tell which half is current, and here the two halves are far
enough apart that you would believe the first one.

The reasoning for keeping it warm is sound and already argued beside the
service block: Fly's proxy gives an auto-started machine ~8s to bind and
this process needed ~10s after LiteLLM landed. Scaling to zero is the
shape this worker wants; it is not the shape it has, and the header now
says so.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@Sbussiso
Sbussiso merged commit a126a07 into master Sep 14, 2026
12 checks passed
@Sbussiso
Sbussiso deleted the docs/agent-is-not-scale-to-zero branch September 14, 2026 04:49
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant