High iowait on n1-highcpu-4 instance

[Eran Sandler]

Hi,

I have an n1-highcpu-4 instance running RabbitMQ. Every couple of minutes it queues about 5k message and have a set of worker machines dequeue things and perform work.

I have very unpredictable iowait times which sometimes hogs the whole machine.

Is it a bug? Is there a way to increase IO operations (maybe software RAID with a bunch of disks like in AWS)?

Would love to get some feedback and help.

Thanks,

Eran

[Eran Sandler]

Thanks for the info.

It's the default 10GB.

I've placed some monitoring on it and it doesn't do more than 50 IO/sec (although I may just be seeing the max IOs it is doing due to its size)

Do you think attaching an additional 100GB drive should be enough? I can try attaching an additional disk of something like 100GB and move RabbitMQs storage to it and see if I get high IO waits again.

Perhaps there is a need for a quick calculator to show the increase in I/O per disk size up to the max I/O per instance type.

Eran

On Mon, Dec 16, 2013 at 12:17 PM, Jens Kuehlers <kueh...@google.com> wrote:

Hi Eran,

how big is the disk attached to the instance? Performance increases with higher disk size.

You can read more about this on this link:

https://cloud.google.com/developers/articles/compute-engine-disks-price-performance-and-persistence

Thanks,

Jens

--
© 2013 Google Inc. 1600 Amphitheatre Parkway, Mountain View, CA 94043
 
Email preferences: You received this email because you signed up for the Google Compute Engine Discussion Google Group (gce-dis...@googlegroups.com) to participate in discussions with other members of the Google Compute Engine community and the Google Compute Engine Team.
---
You received this message because you are subscribed to the Google Groups "gce-discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gce-discussio...@googlegroups.com.
To post to this group, send email to gce-dis...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gce-discussion/aef09cc1-9438-42c8-9018-aa1d636b0a85%40googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--

Jens Kuehlers | Cloud Platform Solutions Engineer | kueh...@google.com 

Google Germany GmbH, Unter den Linden 14, 10117 Berlin
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg
Geschäftsführer: Graham Law, Christine Elizabeth Flores

[Eran Sandler]

Yeah, saw that after I sent the email :-)

I'm trying 200GB for now and will keep on measuring if it fits my workload.

I am a long time AWS user in the uncharted waters of GCE trying to kick it around and see how well it works :-)

Eran

On Mon, Dec 16, 2013 at 12:56 PM, Jens Kuehlers <kueh...@google.com> wrote:

Hi Eran,

currently we suggest a 500 GB disk, but not less than 200 GB for most workloads. The limits are listed on the page I mentioned - it depends very much on the performance the application needs.

Look for the table with "IO Pattern" and "Limit for 1 TB Volume" on the page I mentioned. Since performance increases linearly, you should be able to choose the right size of the disk easily.

Jens

Jens Kuehlers | Cloud Platform Solutions Engineer | kueh...@google.com | +49 30 303986377

[Jens Kuehlers]

Hi Eran,

currently we suggest a 500 GB disk, but not less than 200 GB for most workloads. The limits are listed on the page I mentioned - it depends very much on the performance the application needs.

Look for the table with "IO Pattern" and "Limit for 1 TB Volume" on the page I mentioned. Since performance increases linearly, you should be able to choose the right size of the disk easily.

Jens

Jens Kuehlers | Cloud Platform Solutions Engineer | kueh...@google.com | +49 30 303986377

[Jens Kuehlers]

Hi Eran,

how big is the disk attached to the instance? Performance increases with higher disk size.

You can read more about this on this link:

https://cloud.google.com/developers/articles/compute-engine-disks-price-performance-and-persistence

Thanks,

Jens

On Mon, Dec 16, 2013 at 8:59 AM, Eran Sandler <er...@sandler.co.il> wrote:

--

© 2013 Google Inc. 1600 Amphitheatre Parkway, Mountain View, CA 94043
 
Email preferences: You received this email because you signed up for the Google Compute Engine Discussion Google Group (gce-dis...@googlegroups.com) to participate in discussions with other members of the Google Compute Engine community and the Google Compute Engine Team.
---
You received this message because you are subscribed to the Google Groups "gce-discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gce-discussio...@googlegroups.com.
To post to this group, send email to gce-dis...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gce-discussion/aef09cc1-9438-42c8-9018-aa1d636b0a85%40googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.