Hi all,
we monitor 35 nodes from one master which creates about 6508 graphs.
Today, in graph_strategy=cron the time slot is exceeded (ie. graph generation takes more than 5 mn !) so we have to use graph_strategy=cgi.
But, in our case, cgi generation is very slow. I've traced the code and the bottleneck is in datafile read function which has very poor performance with big datafile (55875 lines in our case). Remember that the datafile is (re)readed for each graph generation !
So, I think about a kind of "cache" to speed up this process.
Please see the attached patch, which is in production in our environment.
Please let's me know what you think about this.
I'm remain available to help you correct this performance problem.
Alex.