I have three different disk pratition on my linode...

AIUI, partitions on Linodes are allocated as LVM logical volumes on the underlying host's RAID-10 array. It's possible, however, that some volumes may be more fragmented than others, or that the host's load balancing algorithms may cause odd results.

@bdonlan:

AIUI, partitions on Linodes are allocated as LVM logical volumes on the underlying host's RAID-10 array. It's possible, however, that some volumes may be more fragmented than others, or that the host's load balancing algorithms may cause odd results.

so what could be a good way to monitor the I/O read writes on our linode?

If you're looking to monitor the amount of I/O being performed, the sum of all three would be the "total". From your kernel's perspective, they are three distinct block devices.

@hoopycat:

If you're looking to monitor the amount of I/O being performed, the sum of all three would be the "total". From your kernel's perspective, they are three distinct block devices.

cacti is widely used but I don't find any topic here on how to do that.

every template I founded doesn't let me do this.

I haven't used cacti (went straight from munin to graphite/statsd), but you might be able to get it to do some sort of overlay, like munin does:

~~http://munin.sodtech.net/hoopycat.com/f … ps-day.png">http://munin.sodtech.net/hoopycat.com/framboise/diskstats_iops-day.png" />

The downside of representing total I/O as a sum of all disk block devices is that it breaks down when dealing with LVM and/or software RAID. On this machine, I have three physical disks (sda, sdb, and sdc) and three LVM logical volumes scattered across them (backuppcdata, root, swap1):

~~http://munin.sodtech.net/hoopycat.com/h … ps-day.png">http://munin.sodtech.net/hoopycat.com/hennepin/diskstats_iops-day.png" />

The sum of all six of those lines would be meaningless. So, maintaining separate graphs for each block device (or at least separate plots on the same graph) is likely the most consistent and intuitive way to handle this across all platforms.~~~~

With munin I'm able to do that but I prefer to use cacti and I'm not able to solve the problem on cacti.

no one uses cacti here to monitor I/O?

there are some guide here on how to use cacti on linode but there is no way to get real I/O data.

http://library.linode.com/server-monitoring/cacti

is there someone from linode who can tell me if my three patitions are on different drives please?

@sblantipodi:

is there someone from linode who can tell me if my three patitions are on different drives please?
No, they are effectively all on the same drive – although that 'drive' is actually a disk array, for speed and redundancy. bdonian was correct when he wrote

> AIUI, partitions on Linodes are allocated as LVM logical volumes on the underlying host's RAID-10 array. …

@pclissold:

@sblantipodi:

is there someone from linode who can tell me if my three patitions are on different drives please?
No, they are effectively all on the same drive – although that 'drive' is actually a disk array, for speed and redundancy. bdonian was correct when he wrote

> AIUI, partitions on Linodes are allocated as LVM logical volumes on the underlying host's RAID-10 array. …

no idea on how to monitor this LVM logical volumes on cacti?

http://docs.cacti.net/usertemplate:data:host_mib:diskio

@iml:

http://docs.cacti.net/usertemplate:data:host_mib:diskio

I'm using that templates since more than a week but there is now way to merge the xvda and xvdc paritions with that templates.