| *** mhen_ is now known as mhen | 02:24 | |
| *** ministry is now known as __ministry | 08:11 | |
| guesswhat[m] | Is anyone using Cinder backend for Glances? I can not get it work. Need some help. Thanks | 09:15 |
|---|---|---|
| admin1 | guesswhat[m] , i guess in real prod, that is counterproductive | 09:56 |
| guesswhat[m] | admin1: well, this is for small setups, maybe even usefull for single node / devstack, where you can allocate disk for glances and cinder | 09:57 |
| admin1 | you can easily do this .. setup a NFS in the same server, and then create 2 folders and use one for glance and one for cinder | 09:57 |
| admin1 | this is the most efficient way to use resources | 09:57 |
| admin1 | this way, you have 1 simple nfs doing 2 folders . so its already complaint with all the latest expectation from openstack ( that cinder and glance is separate ) and on the backend, it will allow you to manage ( if its growing, you can mount something in the backend or just ln to a new folder ) | 09:59 |
| admin1 | and also open you possibility to change glance or cinder to something else .. .. cinder you can have many ( not multiple ceph as nova supports only 1), but you can switch glance to something else as well and just copy the files back | 10:00 |
| guesswhat[m] | @admin1 so nfs is better in terms of resources than for example lvm ? | 10:41 |
| guesswhat[m] | or is NFS faster than LVM ? | 10:43 |
| admin1 | if you are working in constrained resources but still want to be future proof, i would highly suggest to get started with nfs | 10:51 |
| admin1 | as its just 1 daemon to run and whitelist the storage range for the export folders | 10:51 |
| admin1 | faster is subjective .. in terms of bandwidth and the storage used .. i have worked with nfs that gave 500k iops vs lvm that where nowhere nearer | 10:52 |
| guesswhat[m] | ah, thanks, seems that glances does not support nfs directly, but it supports filesystem backend, i guess thats it, right ? | 10:55 |
| matiasbjorling | Hi - I have a bunch of disks (sata) in a nova node, where I would like them to be passed raw into various instances (running on that node). Is there a nice way to do that? | 10:56 |
| guesswhat[m] | in the end, both would be on the same hw, the difference is ceph would consume another resources | 10:56 |
| *** seebaer is now known as seba | 23:19 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!