Harden reinstall_init and add post-saltify readiness gate

- setup/so-functions: dump pre-reinstall salt state (systemctl /
  journalctl / ls /etc/salt / .rpmnew diff) to the setup log so a
  failed reinstall leaves a usable post-mortem; swap the manual
  rm -rf of /etc/salt/* for `dnf -y remove salt` so package configs
  get cleaned up properly.
- setup/so-setup: replace the `sleep 2 / state.show_top / sleep 2`
  dance after saltify with a readiness gate that waits for
  /etc/salt/pki/master/master.pub, runs check_salt_master_status,
  and then wait_for_minion_key_pending before salt-key -ya. Fixes
  reinstalls on 3.x timing out on "Unable to sign_in to master".
- salt/common/tools/sbin/so-common: add wait_for_minion_key_pending
  helper, polls `salt-key -l pre` until the minion appears.
This commit is contained in:
Mike Reeves
2026-04-23 17:42:53 -04:00
parent 8b2064f1c1
commit cab57edf7c
3 changed files with 48 additions and 4 deletions
+11 -3
View File
@@ -724,10 +724,18 @@ if ! [[ -f $install_opt_file ]]; then
# Install salt
saltify
check_sos_appliance
# Wait for salt-master to be actually running and have its PKI
# ready after a fresh saltify. Without this, salt-key operations
# silently race the daemon and the key accept no-ops, which is
# what was causing reinstalls on 3.x to hang on state.show_top.
retry 30 2 "test -f /etc/salt/pki/master/master.pub" \
|| fail "salt-master did not initialize PKI after saltify"
check_salt_master_status \
|| fail "salt-master not accepting calls after saltify"
logCmd "salt-key -yd $MINION_ID"
sleep 2 # Debug RSA Key format errors
logCmd "salt-call state.show_top"
sleep 2 # Debug RSA Key format errors
wait_for_minion_key_pending "$MINION_ID" 30 2 \
|| fail "salt-minion never presented its key to salt-master"
logCmd "salt-key -ya $MINION_ID"
logCmd "salt-call saltutil.sync_all"
# we need to sync the runner and generate the soqemussh user keys so that first highstate after license created