Re:One Million Database Connections
by aarondf on Tuesday November 01, 2022 @02:41PM
(#33425568)
This is a good call out.
Data corruption is always possible due to software, firmware, hardware bugs and failures... but that’s not specific to OOMs.
You could have non-crash safe settings like sync_binlog != 1 or innodb_flush_log_at_trx_commit != 1, but with the default settings of MySQL 8.0 it’s entirely crash safe in every way (binlog events etc). I think that part of the post needs to be updated. It was a bit of an off-handed comment but it's not as clear and accurate as it should be.
I’m going to replace `potential data corruption` with `potential downtime` as it could initiate a failover and the process will have to restart and go through crash recovery which can take some time.
Data corruption is always possible due to software, firmware, hardware bugs and failures... but that’s not specific to OOMs.
You could have non-crash safe settings like sync_binlog != 1 or innodb_flush_log_at_trx_commit != 1, but with the default settings of MySQL 8.0 it’s entirely crash safe in every way (binlog events etc). I think that part of the post needs to be updated. It was a bit of an off-handed comment but it's not as clear and accurate as it should be.
I’m going to replace `potential data corruption` with `potential downtime` as it could initiate a failover and the process will have to restart and go through crash recovery which can take some time.