Mongodump on Gridfs is killing the host IOs
Posted
by
Raphael
on Server Fault
See other posts from Server Fault
or by Raphael
Published on 2014-05-29T20:44:43Z
Indexed on
2014/05/29
21:33 UTC
Read the original article
Hit count: 280
I'm trying to make a mongodump from our production mongodb while the production is running.
We have three production instances, one regular mongodb, one with very few gb of data on gridfs, one with a larger amount of data on gridfs. All mongodb instances are running in version 2.4.9 on a ubuntu 10.04 virtual server.
I use a mongodump command to export the bases to another server.
Unfortunately our machines are virtually hosted in a "low performances" datacenter (vmware based) so when I try to export the large gridfs db, the disk IO hits 100% (and 50% of the cpu starts waiting for IO too).
This has a very negative impact on the production applicatiosn because db access time is excessively increased, making the applications unusable.
I'm looking for a way to regulate the mongodump so the export goes slower but cooler on the hardware ressources allowing better performances for the applications to run.
Has anyone had a similar scenario ?
© Server Fault or respective owner