Sorry this is a bit off-topic, but maybe someone has some insight.
I have a pretty big cluster with a lot of shards (26618).
I am not following good practices as all data nodes are also master.
That disclaimer out of the way let me discribe the issue: wheneverf I try to reload or create an index-pattern in kibana, the cluster goes haywire for half an hour, load goes up to 1, garbage collector maxes out, heap usage as usual. Eventually the operation fails, and my logs say a lot of gc overhead messages.
does anyone know what could be wrong ?