I tried to reproduce an issue we have on a customer environment.
My goal was to reproduce a LOCK WAIT TIMEOUT when 20 instances insert a row into the same SPIDER/TOKUDB table but I'm not able to reproduce it for the moment. A maximum of 100,000 rows can be inserted each day.

However, I have this issue about some duplicate entry for PRIMARY KEY.
The PRIMARY KEY is composed by an AUTO-INCREMENT and a date and after 24000 insertions, the process failed several time with a PRIMARY KEY issue.
I'm able to reproduce it:

• run the ./run_20_instances (10 on 1 spider node and 10 on the other)
• wait about 30000 insertions

I updated the script to reproduce the queries flow from our tools on 2 tables, one is use to organize a list of actions, the other is use to log each events from the process of the action. So one is use as a PULL/PUSH table and the other receive new lines.
With this script, I'm able to reproduce a TIMEOUT (and also Duplicate key) on a Spider system but not on a TokuDb system.

• run the ./run_20_20_instances
• wait about 30000 insertions

I don't think this issues are linked to my issue because I can't reproduce it without a Spider system but it shows that the SPIDER engine can be instable.