I’m encountering an issue on an Amazon RDS box running mysql 5.6.
There’s a application process that runs lots of concurrent single-row updates on an innodb table, using the primary key.
update a set blah=blah … where ID = X;
We’ve been consistently seeing 8-10 threads stuck in a commit state for 2-8 seconds.
This RDS instance is dedicated to this one schema, and this is the only table receiving any significant write activity during this time.
Table has approx 70k records.
buffer_pool=2.7gb
Row_format=compact
file_format=antelope
The table has 3 text columns and one mediumtext column, which seems significant based on what I’ve googled.
We updated the server to run file_format=barracuda and row_format=compressed
Still get similar behavior, but now the 8-10 threads are stuck in “query end” state for several seconds at a time.
What should I check for next?