fix(logs): use PSR log level when using PSR-3 logger #1907
Conversation
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
| return \Psr\Log\LogLevel::CRITICAL; | ||
| case 'info': | ||
| default: | ||
| return \Psr\Log\LogLevel::INFO; |
There was a problem hiding this comment.
Not sure what's a good default level here but also this shouldn't really occur
|
The one existential question I have is what does it mean if Sentry is logging a fatal error successfully? Does that mean Sentry is working normally - i.e. it's not really e.g. critical? :) |
|
@mfb could you clarify your message above, as it raises some questions we do not fully understand. |
|
If you look at HttpTransport, it logs at info level when an item is sent (whether debug or fatal); warning if rate limit is hit; error if sending fails. I was referring to how you're using a different philosophical approach with SDK logging for log items, by choosing a level based on the payload rather than how the SDK is functioning. |
| // We check if it's a `LogsLogger` to avoid a infinite loop where the logger is logging the logs it's writing | ||
| if ($sdkLogger !== null) { | ||
| $sdkLogger->log((string) $log->getLevel(), "Logs item: {$log->getBody()}", $log->attributes()->toSimpleArray()); | ||
| $sdkLogger->log($log->getPsrLevel(), "Logs item: {$log->getBody()}", $log->attributes()->toSimpleArray()); |
There was a problem hiding this comment.
What (I think) @mfb points out is that we copy the log level from the log item instead of logging always as "DEBUG" or "INFO" and mentioning the log level of the log item in the log message.
This was introduced by me I think, this maybe should be updated to mention the log item level in the log message and set our log level to debug (?).
In addition, the comment above this if is outdated and incorrect :) The LogsLogger never made it in but the comment remains 🙈
There was a problem hiding this comment.
https://develop.sentry.dev/sdk/telemetry/logs/#other this is by design.
There was a problem hiding this comment.
In that case I did this very by the book 😎
Then maybe the comment could be updated but no other changes are needed.
There was a problem hiding this comment.
Ok, great - as long as it's by design :)
It's not 100% clear to me but I guess "appropriate" is saying the Sentry log level should be translated to an appropriate level, e.g. log a "fatal" to the console as "critical". I'd also find it reasonable to just print the "fatal" level as part of an "info" log message - either way works for me.
Fixes an issue where we use a PSR-3 logger in the
LogsAggregatorbut pass SentryLogLevelinto it, which might lead to errors in different loggers as they expect PSR-3 levels.Closes #1902