johnnyq
|
bce68aac6a
|
Add a collation check and repair to Admin > Debug: functions/database_collation.php reads from db.sql each table's default collation and each text column's, compares them with information_schema and can give every table that differs the collation db.sql has for it, one ALTER a table with a MODIFY a column built from its SHOW CREATE TABLE line with only the collation replaced, as database 3.6.5 does, so a binary column that was flattened goes back to utf8mb4_bin too and no column type changes; the ITFlow section gains a Database collation check that warns when any table differs, and the Database card lists them under Structure a row a table, in Copy for support as well, with a Repair collations button behind a confirm; the button posts to the new admin/post/debug.php, CSRF-checked and not tied to the request so a large table is not cut off with the browser, which writes an audit log entry naming the tables changed and a warning in App Logs for each one it had to leave, a table being changed whole or not at all; tables and columns db.sql does not have are not looked at; run through the page and against the whole schema on MySQL 8.0 and MariaDB 10.11; add both to the 26.11.7 changelog section
|
2026-11-11 20:33:19 -05:00 |
|
johnnyq
|
87cdf79b28
|
Make managed mode off by default and turned on per side in config.php - $config_enable_managed_instances = true; adds Admin > Managed Instances, the api/managed endpoints and the Managed Instances cron job, and $config_enable_managed_by = true; (or $config_manager_required = true;) adds Admin > Managed By, the Manager Check-in cron job and support sign-in; with a side off its menu entry, pages, modal, post handler and cron job are gone, the manager endpoints answer 503 so instances keep their connection and retry, a connected instance stops checking in and its locks, read-only and suspended lift while its connection is kept, any open support session is ended, and the support user still takes no seat; new managerEnabled(), managedEnabled() and managedConnected(), the cron registry and health check leave out the jobs of a side that is off, changelog updated
|
2026-11-05 20:17:42 -05:00 |
|
johnnyq
|
fb9534b0a2
|
Rebuild Admin > Debug as a system health page: problems and warnings first with what to do and where, key figures along the top, checks by section (ITFlow and database version, updates, cron and jobs, mail queue and email to tickets, backups, PHP version, extensions and limits, HTTPS, certificate, base URL and HTTPS only, folders and disk space, server and clock drift, errors logged), database size, biggest tables and structure compared with db.sql ignoring integer display widths, uploads by folder, and a Copy for support plain-text report without host names, paths or keys; row counts use the database's estimate with a Count exactly option, the uploads scan stops after a few seconds, the certificate is checked against config_base_url instead of the request host, HTTPS counts X-Forwarded-Proto, every query degrades instead of failing when the database is behind the code, and the page no longer closes the database before the footer; new functions/system_health.php
|
2026-10-24 14:46:01 -04:00 |
|