You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
fix: emit add-host at podman pod create instead of podman run
Current podman rejects --add-host on a container joining a pod
("network cannot be configured when it is shared with a pod").
Moves alias/hostname add-host emission from run_flags to
pod_create_flags, alongside dns/sysctls, and folds extra_hosts in as
a pod-level key merged into the same add-host set (conflict refused
rather than guessed at).
summary: Move all `--add-host` emission (alias/hostname resolution and `extra_hosts`) from per-service `podman run` to `podman pod create`, since current podman rejects network config on a container joining a pod; `extra_hosts` becomes pod-wide like `dns`/`sysctls`, with a same-host-different-address conflict (against another service's `extra_hosts` or an alias's fixed `127.0.0.1`) refused rather than guessed at.
3
+
---
4
+
5
+
# Design: pod-level `--add-host`
6
+
7
+
## Summary
8
+
9
+
Current podman rejects `--add-host` on `podman run` for a container joining a
10
+
pod (`invalid config provided: extra host entries must be specified on the
11
+
pod: network cannot be configured when it is shared with a pod`), discovered
12
+
by the new integration harness (`planning/changes/2026-07-12.04-integration-harness.md`)
13
+
on its first real CI run. Every generated script with more than a trivial
14
+
single-service pod currently fails at run time. The fix hoists all
15
+
`--add-host` emission onto `podman pod create`, alongside `dns`/`sysctls`
16
+
(`compose2pod/pod.py`). This forces `extra_hosts` — previously genuinely
17
+
per-service — to become pod-wide like `dns`/`sysctls` already are; a same-host
18
+
conflict is refused rather than resolved by guessing.
0 commit comments