rdmpw3275m-changelog-20260921-1422-issue-20260921-04-redeploy-daemon-restart

Fixed false daemon_restart reporting and unnecessary daemon restarts in redeploy_app_fleet.zsh.

Scope

• Hosts: rdmsm4x, rdmbair15m5, rdmpw3275m
• Projects: Tyrell, issues

Summary of Changes

• Addressed ISSUE-20260921-04: during redeploy_app_fleet.zsh --execute-canary, install_service.zsh was executed unconditionally on every host, booting out and restarting the daemon even when the running process was already executing the exact candidate release. This silently terminated in-flight long-running memory soaks (such as the Build 30 six-hour physical-footprint gate).
• Additionally, verify_tyrelld_process.zsh was run immediately following install_service.zsh; because the candidate had just been bootstrapped, verify_tyrelld_process.zsh saw no need for its own kickstart and printed daemon_restart=not_required, masking the restart that install_service.zsh had just performed.
• Updated scripts/redeploy_app_fleet.zsh (v3.3):
  1. Pre-probes the host via verify_tyrelld_process.zsh --expected-sha256 "$daemon_hash". If the candidate daemon is already verified running and healthy, skips install_service.zsh and avoids killing the running process unless --reinstall-daemon is explicitly passed.
  2. Captures prior_daemon_pid before installation. When install_service.zsh runs, filters out misleading daemon_restart=not_required to emit daemon_restart=performed, and prints a loud warning that in-flight soaks against the prior PID are void.
  3. Added --reinstall-daemon CLI option to allow operators to force daemon re-staging and service reload when desired.
• Updated Tests/ScriptTests/tyrelld_process_restart_contract_test.zsh with assertions covering pre-candidate probing, daemon_restart=performed reporting, voided soak warning, and --reinstall-daemon support (20/20 passed).
• Merged to canonical main in apps/Tyrell (commit 7285b0b), resolved ticket in dev/issues (commit 3573fbf), and pushed both repos to fleet and backup.
• Broadcasted issue-20260921-04-resolved via agent_msg.zsh.

Verification Evidence

• zsh Tests/ScriptTests/tyrelld_process_restart_contract_test.zsh: 20 passed, 0 failed.
• zsh Tests/ScriptTests/app_deployment_plan_contract_test.zsh: passed.
• zsh Tests/ScriptTests/app_operational_entrypoints_contract_test.zsh: passed.
• zsh Tests/ScriptTests/applications_deploy_contract_test.zsh: passed.
• git push fleet main && git push backup main in apps/Tyrell and dev/issues: clean.

How to Undo

• git -C ~/dev/apps/Tyrell revert 7285b0b
• git -C ~/dev/issues revert 3573fbf