Percona MySql version is :
mysql Ver 14.14 Distrib 5.5.46-37.5, for debian-linux-gnu (x86_64) using readline 5.1
OS Version is :
Debian 7.9
MySql master has a separate tmp partition, of 100Gb in size on an SSD. Mysql itself is on an 900Gb partition, again on SSD. BinLogs to a separate non-ssd partition, of around 500Gb. Each partition had space.
Someone ran a bad query, select count(*) ......., and then aborted it. The logs that came after cancelling the sort is in the below pastebin link:
The abort caused mysql to throw an error "during flush stage of the commit" and since the variable binlog_error_action was set to IGNORE_ERROR (my fault, should have been ), mysql stopped writing to binlogs and all slaves stopped replicating.
My Question is: why did an error came "during flush stage of the commit" and why did it affect the binlogs?