Ask Your Question
0

Switch speed for instances on shared storage

asked 2015-01-26 04:32:03 -0500

DaveJ gravatar image

Hi,

I'm looking at requirements for running instances on shared storage, potentially Ceph. I'm assuming at least a 10G switch would be required to support this?

If I was using Ceph, running on compute hosts with a replica count of 2, would it be possible to configure Ceph and Openstack in such a way, that the instance data was always served up from the local disk if was available, and only fetch from the remote disk in the event of failure.

The use-case here is to be able to evacuate/re-launch instances when a compute host completely fails and I loose the local disks.

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
0

answered 2015-01-26 06:41:14 -0500

The size of your network connectivity really depends on the scale of your compute environment, Ceph cluster and your rate of change against that Ceph cluster which will increase the amount of replication between the Ceph OSD nodes. 10GB might work on a small cluster/compute environment but as your environment grows, you may find that you are saturating your backend switches. One way to prepare for scale would be to ensure you have your replication traffic on your Ceph cluster on a dedicated 10GB link with a physically separate switch from the 10GB front side client traffic to the OpenStack compute nodes. Ensure your switches can support up to 40GB if again you plan on growing the environment very large.

With regards to your question about configuring Ceph as a standby in the event that the local disks on a compute node are not available, I have not seen anything that would support this configuration. First, the local disk as it was changing for a compute instance would somehow need to be replicating from a block level over to the Ceph cluster to ensure any writes were kept in sync from the local disk to the Ceph disk. Along with that, OpenStack would have to have some kind of understanding/configuration that indicates that when local compute disk are down request storage from the Ceph cluster. Niether of these configuration items exist today.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2015-01-26 04:32:03 -0500

Seen: 166 times

Last updated: Jan 26 '15