Skip to content

Scheduler start after an update, and firewall rejections in the connection diagnosis - #943

Merged
twikles merged 3 commits into
testfrom
fix/schema-changed
Nov 15, 2026
Merged

twikles merged 3 commits into
testfrom
fix/schema-changed

Conversation

@twikles

@twikles twikles commented Nov 15, 2026 •

Copy link
Copy Markdown
Member

Both seen on the production node after 1.27.2:

  • hyperlite-scheduler failed at its start at nearly every update since 14/11 ('database schema has changed' in _attach_access, restarted by systemd 5 s later). get_conn retries with a new connection.
  • The connection diagnosis left a rejected try ('No route to host', firewalld on the target) undecided; it now says the target's firewall, before an empty port.
    Tests: tests/test_console_27.py (13 passed) + database/cfs tests; vitest 125.

Each service replaces the replication triggers at its start, so at every update one of them reads the schema while the
other changes it: hyperlite-scheduler stopped with 'database schema has changed' at nearly every update since 14/11
(systemd restarted it). get_conn now opens a new connection a moment later, up to five times.
…rewall

Seen on the production node: a try to a port the guest's firewalld rejects answered 'No route to host' and the verdict
stayed undecided.
@twikles twikles changed the title Open a connection again when another service changed the schema Scheduler start after an update, and firewall rejections in the connection diagnosis Nov 15, 2026
@twikles
twikles merged commit 1007aae into test Nov 15, 2026
3 checks passed
@twikles
twikles deleted the fix/schema-changed branch November 15, 2026 12:52
@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