Skip to content

Fix the schema update on a busy database and VM addresses for diagnosis and remote desktop - #938

Merged
twikles merged 1 commit into
testfrom
fix/1.27.1
Nov 15, 2026
Merged

twikles merged 1 commit into
testfrom
fix/1.27.1

Conversation

@twikles

@twikles twikles commented Nov 15, 2026

Copy link
Copy Markdown
Member

Found while checking 1.27.0 on the production node:

  • hyperlite-scheduler failed once at start (sqlite3.OperationalError: database is locked in install_triggers), restarted by systemd. init_db now retries.
  • The connection diagnosis had no address for either VM (API's read-only libvirt: read only access prevents virDomainQemuAgentCommand), so no live try. New guest_agent.address_now falls back to the DHCP lease; also used by rdp.open.

Tests: tests/test_console_27.py (12 passed) and the database tests. Remote desktop itself verified for real on antho against xrdp in a disposable VM (session opened, screen read through the relay).

…osis and remote desktop

- init_db tries the schema update again (up to six times) when SQLite answers 'database is locked': replacing the
  replication triggers needs the database alone, and the API's own writes made hyperlite-scheduler fail at its start
  after the update to 1.27.0 (systemd restarted it).
- guest_agent.address_now: the agent's answer waited for, else the DHCP lease. The API's read-only libvirt cannot ask
  the agent, so the diagnosis had no address and made no live try; the remote desktop could miss the address on a
  first call.
@twikles
twikles merged commit 5f347cb into test Nov 15, 2026
3 checks passed
@twikles
twikles deleted the fix/1.27.1 branch November 15, 2026 12:01
@twikles twikles mentioned this pull request Nov 15, 2026
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