Testing Disk Read / Write on Host Node

[Tyler]

Does anyone know of a shell script for testing disk read/write speeds? I am trying to find a way to ensure that my read/write speeds stay normal.

I had a few ideas for this:

• Shell script that runs a DD test and emails if the test comes back with a speed below X MB/s -- then run this script on a cron job every few days. 
• Run either a script or daemon to monitor I/O wait and emails if the wait comes back as high

I'm thinking option #2 would probably be better since it won't hammer the disk for a few seconds simply to test (not that it is particularly significant, but in a production machine, you would want to eliminate unnecessary writing as much as possible--it's also compounded by the fact that a cron job would be run multiple times in option #1).

I'm not of the technical nature, so if you know of a better way to do this, please let me know!

I have no qualms with paying to have a script written for this, I would obviously like to know if there is something in the open source community that already exists for this need. Maybe there is a way to do this with Nagios - but I want it to send an email (so it can be piped to an internal admin queue). Not too sure.

 

[AshleyUK]

Does anyone know of a shell script for testing disk read/write speeds? I am trying to find a way to ensure that my read/write speeds stay normal.

I had a few ideas for this:

• Shell script that runs a DD test and emails if the test comes back with a speed below X MB/s -- then run this script on a cron job every few days. 
• Run either a script or daemon to monitor I/O wait and emails if the wait comes back as high

I'm thinking option #2 would probably be better since it won't hammer the disk for a few seconds simply to test (not that it is particularly significant, but in a production machine, you would want to eliminate unnecessary writing as much as possible--it's also compounded by the fact that a cron job would be run multiple times in option #1).

I'm not of the technical nature, so if you know of a better way to do this, please let me know!

I have no qualms with paying to have a script written for this, I would obviously like to know if there is something in the open source community that already exists for this need. Maybe there is a way to do this with Nagios - but I want it to send an email (so it can be piped to an internal admin queue). Not too sure.

I use NewRelic, may be a bit overkill for your needs, but provides full monitoring of all levels of the node including I/O and allows you to setup custom alert levels and allows you to look back at past data over a set period, and has a good Mobile App and reporting system.

Places very minimal "overhead" to the node.

 

[MightWeb-Greg]

I use NewRelic, may be a bit overkill for your needs, but provides full monitoring of all levels of the node including I/O and allows you to setup custom alert levels and allows you to look back at past data over a set period, and has a good Mobile App and reporting system.

Places very minimal "overhead" to the node.

I'll check this out myself. I just stood up a new node myself and I seem to have problems with read/writes. It's running at 3gb/s speeds vs the 6gb/s it should be.

I'm just running a dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync

But this may give me a better idea as to what is going on in the future.

 

[Navyn]

We tried New Relic with some of our vps clients they reported that it uses high resources.

But not tested on node servers yet.

Hope it will work fine for you.