In the past couple of weeks I have been working on implementing an elasticsearch solution, combined with logstash we hope to implement it to replace the existing Splunk system that exist within the infrastructure.
I have build a Chef cookbook to implement it, and with in the confines of the testing it worked, then one of the developers “complained” that there is a newer version and as they use it in their environment we should use it too.
Suffice to say that I had to spend the entire day trying to fix it in the system we are attempting to have in production, only to eventually learn that the versions are incompatible for the method we want to use and then I had to roll back all the version and re-apply the system.
In many places I have witnessed a mentality were the developers are setting the standard policy on tools and implementation – which more often then not usually backfires when the lead developer leaves or a system admin gets “dumped” with a tool and is expected to know it from day 1, and when it breaks, no one asked the developers anything and the blame falls on the ops person.
So today when the developer put his input about upgrading the component that was incompatible, instead of rolling back to a version we know that works, I told him flat out – No, and only when we System decide we want to upgrade – we will do it – not on their wishes.