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 etcd static pod hostname renaming for non-shell containers
Required due to openshift/cluster-etcd-operator#1606
The `etcd-ensure-env-vars` init container changed from an inline bash
script to a Go binary (cluster-etcd-operator ensure-env), which caused
`fix_etcd_static_pod_container` to skip renaming hostname references in
command args. The function relied on finding a `"#!/bin/sh\n"` prefix to
identify args containing hostname-dependent strings, and silently
skipped the entire replacement block when no such prefix was found.
This left stale `NODE_{old_hostname}_*` references in the container's
command args while the env var definitions were correctly renamed,
causing etcd to fail at startup with "environment variable
`NODE_<old>_ETCD_URL_HOST` must be set".
Apply hostname replacements to all command args unconditionally instead
of gating on the shebang heuristic.
0 commit comments