Ssl_ctx_verify_depth returns the value of the ssl library using the SSL_get_verify_depth https://www.openssl.org/docs/manmaster/man3/SSL_get_verify_depth.html where -1 is wrapped to the maximum. It does however mean the SSL default depth. It won't reach that high as you won't place an arbitrarily length CA chain on the server.

The wsrep values look like an arbitrary number, however once you enable galera these will have much saner looking values.

Thanks so much for the info, Daniel. That makes perfect sense.

Sorry, reopening this, because we are sometimes seeing strange values for other variables, like Innodb_os_log_pending_writes. Sometimes the value is 0, other times it is 18446744073709552000. For an unsigned 64-bit integer, that's -384. What could explain that?

The status variable Innodb_os_log_pending_writes was removed in MDEV-14425 (MariaDB Server 10.8.0). Another user is reporting seeing negative or wrapped values for it in 10.5.15 but not 10.3.28.

A possible explanation for the wrap-around of InnoDB_os_log_pending_writes could be that srv_stats.os_log_pending_writes is essentially declared as size_t, not std::atomic<size_t>. So, race conditions between increment and decrement operations are possible.

Theoretically, we could ‘fix’ this by always reporting the status variable as 0 in releases older than 10.8, but then someone could be upset about it. I think that using proper std::atomic declaration is out of the question, because it would very likely introduce performance degradation.