Side effect of decreasing index.translog.flush_threshold_size


In my test on my test cluster, after changing index.translog.flush_threshold_size from default 512MB (default) to 32MB, old gc occurs much less frequently. (I think this is because that the buffer is enoughly small, so that free'd Trans Log objects are collected in survivor area before moving to Old Gen.)

So, I will change index.translog.flush_threshold_size in my production cluster. But before that I'd like to know the side effect of small size of index.translog.flush_threshold_size.

I hope someone could help me.

I use ES 2.3.


Having a smaller translog size will lead to more flushes, which means more frequent Lucene commits and thus more IO.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.