Hello,
For sometimes, we were asking ourselves who was due to micro-slowdowns on the server which translate into an IOwait spike, and we decided to conclude it 4 days ago by mobilizing external experts ...
Yesterday, by ourselves, we found the key to undo with the observed logs log, the problem occurs every second 5-6-7-8 every time, so we thought about the cron ... Is a fact that the service cron of the server provided by centos lacks something, ie it executes the cron of the server at the same second whatever the minute, and the cycle of the cron, and it is in The case of a powerful server, even in SSD, and with quite a lot of users ...
After consulting the experts on a solution, the first proposed is a good solution, and it is not certain at this stage that there is another: insert in cron this:
Sleep $ (expr $ RANDOM \% 30);
This dilutes the execution of the cron on one second at random the first 30 seconds ...
As a result, we have modified the visible cron directly in the server's cron files (others are not visible because managed inside the CMS)
We have updated accordingly the article on the cron so that out user tries to do as recommended ...
Sunday, January 15, 2017
