IO spike: how to know if cluster rebalanced


(John Chang) #1

We saw a ~45 minute IO spike in our elastic search nodes today, and during that time, we had no spike in indexing or querying. The only thing we can guess happened was a rebalance. Is there any way we can confirm this theory for sure? What else might cause this spike if our idexing/querying nodes were spiking?

Thanks


(Shay Banon) #2

One other reason for a spike in IO is large merge that happened to a lucene
index. Each lucen index is constructed of several segments, which gets
merged every so ofter (I can provide more details on how this works if
needed). A spike might be caused because a large merge between big segments
has occurred.

-shay.banon

On Thu, Nov 4, 2010 at 10:12 PM, John Chang jchangkihtest2@gmail.comwrote:

We saw a ~45 minute IO spike in our elastic search nodes today, and during
that time, we had no spike in indexing or querying. The only thing we can
guess happened was a rebalance. Is there any way we can confirm this
theory
for sure? What else might cause this spike if our idexing/querying nodes
were spiking?

Thanks

View this message in context:
http://elasticsearch-users.115913.n3.nabble.com/IO-spike-how-to-know-if-cluster-rebalanced-tp1844345p1844345.html
Sent from the ElasticSearch Users mailing list archive at Nabble.com.


(system) #3