My node crawls when I have Elasticsearch store its data on GPFS


(HydraDB) #1

Hello everyone,

I have been struggling with setting up es while having it's data stored on
GPFS. Hopefully, someone can give me some ideas that may rectify the
problem.

The only real setting change that I made was to have
-Des.path.data=/mygpfs/data

When I start es the node crawls and simple file system operations like 'ls'
take minutes to respond. As an experiment, I tried using the local
filesystem /tmp
and the performance was excellent.

I am using

  • elasticsearch-0.19.10
  • Red Hat Enterprise Linux CentOS release 6.2 (Final)

Any ideas would greatly be appreciated. Additionally, if anyone else is
using GPFS, can you tell me which driver set you are using?

Thank you in advance,
cw

--


(Dave) #2

Hello,
Have you had any luck with GPFS? We have installed it and seeing similar bottle necks and lots of shard relocations?


(system) #3