OpenVZ Forum


Home » General » Support » OpenVZ IO Isolation ? (OpenVZ IO Isolation Details)
OpenVZ IO Isolation ? [message #42557] Thu, 28 April 2011 14:22 Go to previous message
andre is currently offline  andre
Messages: 36
Registered: January 2008
Member
Hello,

We have been working with OpenVZ for some years on some servers and have found a problematic behavior with one of our customers.

There are just 2 VE's on this hardware (let's call them VE1 and VE2). Each one with 25% of the resources (to give room for more 2 VEs that would be activated later). It is a Core2Quad with 8GB RAM and RAID1 SATA.

The problem:

VE1 is able to slow down the whole server, making IOWAIT for all VEs really high. To make sure it was not hardware related, we moved the VEs to another hardware (with latest stable kernel: 2.6.18-238.5.1.el5.028stab085.2) and got the same behavior.

Each VE has CPULIMIT=90% and CPUS=1

Reading "IOACCT", read/write/nice/etc information seems pretty "normal" but DIRTY BYTES is high.. So I thought: Could OpenVZ IO scheduler not be counting this ?

I am not a kernel expert (nor have basic kernel coding knowledge), just trying to understand what could cause this and if there are any solutions. At this moment we are keeping VE1 alone at a hardware so it won't impact other customers.

IOWAIT shows VE1's mysqld as the responsible. One VE's process being able to make the whole node near 100% IOWAIT is worrying Sad


Thanks.
 
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Previous Topic: /dev/shm missing from fedora-15-x86 template
Next Topic: NFS on Ubuntu 11.04
Goto Forum:
  


Current Time: Mon Oct 07 09:48:50 GMT 2024

Total time taken to generate the page: 0.05187 seconds