Slow Marvel Response Times

monitoring

(Matthew Iverson) #1

When looking at Marvel I am noticing that my response times have gotten really slow. I have all 3 beats running on a handful of servers and am collecting considerable amounts of data.

Here are some screen shots showing timing when I view my indices in marvel

Here are some other stats about my cluster
Nodes: 5
Indices: 97
Memory: 3GB / 5GB
Total Shards: 887
Documents: 234,149,993
Data: 92GB
Version: 2.3.3

Also I have a dedicated master node with no data. My Kibana instance is connecting to that node. The beats indexes are separated out by beat type and day.

Thanks for any tips!


(Matthew Iverson) #2

So it looks like my instances were using the default 1gb limit for Java. I have upped this to 50% of the server size (4GB per node) and everything is running significantly faster. I thought it used 50% by default but looks like I was wrong


(system) #3