I have a cluster that's encountering memory pressure. It's predominantly
tuned for write performance (logs). Occasionally we get a query sent in
that sorts the entire dataset by timestamp which explodes our fielddata
cache size. It doesn't lead to OOM errors, but a lot of GC churn so we
have to go in and manually clear the fielddata caches. I want to set the
indices.fielddata.cache.size to something (let's say 30% for the sake of
argument). My question is... what will happen if the size of the fielddata
cache required for that query is more than the 30% I've allocated for the
node? Initial tests seem to indicate that things "just work" (I get a
response that looks valid, the caches are kept in check, etc) but I can't
really validate that the results are properly sorted etc (too much data).
I know we really need more memory and/or more nodes. Just thought I'd ping
the experts if they know for sure what to expect....
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to email@example.com.
For more options, visit https://groups.google.com/groups/opt_out.