We are currently experiencing peaks in workers which causes Manticore to lag.
The image below show the result of SHOW STATUS.
Normally the workers_active are between 20-50, but sometimes there is a sudden peak. workers_active, is than rising shortly which causes Manticore to become very slow shortly, and after a few minutes it goes back to normal.
The amount of queries seems pretty stable.
Anyone has an idea what could cause these peaks? Or how to further debug the source of the peaks?
what manticore version do you use?
could you line up your queries from the query.log with the lag case?
do you use only searches or insert / replace / delete or update statements?
“could you line up your queries from the query.log with the lag case?”
What do you mean by this? If there are specific (returning) queries during the lag period?
yes what kind of queries (full text or full scan)?
to what indexes? plain, RT, distributed?
Good point, I will start logging the queries and check which run during the lag.
