Ensure logger creation is consistent between Web and Collector
Create logger in main, pass down to downstream functions (like gin)
In debug mode, print a copy of AppConfig
Better debugging for logger.
When a transaction cannot lock the database, because it is already locked by another one,
SQLite by default throws an error: database is locked. This behavior is usually not appropriate when
concurrent access is needed, typically when multiple processes write to the same database.
PRAGMA busy_timeout lets you set a timeout or a handler for these events. When setting a timeout,
SQLite will try the transaction multiple times within this timeout.
https://rsqlite.r-dbi.org/reference/sqlitesetbusyhandler
retrying for 30000 milliseconds, 30seconds - this would be unreasonable for a distributed multi-tenant application,
but should be fine for local usage.
added mechanism for global settings (PRAGMA and DB level instructions).
fixes#341
update the frontend to persist settings to the database.
Using ScrutinyConfigService instead of TreoConfigService.
Using snake case settings in frontend.
Make sure we're using AppConfig type where possible.
Settings are loaded from the DB and added to the AppConfig during startup.
When updating settings, they are stored in AppConfig, and written do the database.
- updated dbdiagrams schema
- [BREAKING] force failure if `notify.filter_attributes` or `notify.level` is set
- added Settings table (and default values during migration)
- Added Save Settings and Get Settings functions.
- Added web API endpoints for getting and saving settings.
- Deprecated old Notify* constants. Created new MetricsStatus* and MetricsNotifyLevel constants.
This reverts commit 64e1c93d16.
https://gitlab.com/cznic/sqlite does not support Solaris.
> build constraints exclude all Go files in /home/runner/work/scrutiny/scrutiny/vendor/modernc.org/libc/errno
related #120