|
=== dendro-afk is now known as dendrobates |
|
=== Hussain is now known as Guest28163 |
|
=== marrusl_ is now known as marrusl |
|
=== dendrobates is now known as dendro-afk |
|
=== dendro-afk is now known as dendrobates |
|
=== Hussain is now known as Guest21263 |
|
=== dendrobates is now known as dendro-afk |
|
=== dendro-afk is now known as dendrobates |
|
[21:23] <thomi> is it still the case in precise that I can't use juju to deploy to a rackspace cloud instance? |
|
[23:09] <SpamapS> thomi: There's no RS cloud provider been written, no. |
|
[23:09] <thomi> SpamapS: ok, thanks. |
|
[23:11] <nOStahl> hi guys |
|
[23:11] <nOStahl> i got my hard drives in today |
|
[23:12] <nOStahl> two caviar black 1.5 terabyte setup as raid1 in a hp dc5750 tower |
|
[23:12] <nOStahl> I am wondering if I should use openstack or eucalyptus to setup a cloud? |
|
[23:13] <SpamapS> nOStahl: on a single server? |
|
[23:13] <nOStahl> i have dell optiplex 760 two if them for the nodes |
|
[23:13] <SpamapS> nOStahl: I would go with OpenStack. It seems to have the brighter future and is receiving more attention on Ubuntu. |
|
[23:14] <SpamapS> nOStahl: neither of them are very easy to deploy though |
|
[23:14] <nOStahl> I want to be able to have a production instance for web server , email, calendar , contacts servers |
|
[23:15] <nOStahl> and be able to launch an instance for testing stuff etc. |
|
[23:15] <nOStahl> all small stuff max of 50 emails low usage etc |
|
[23:15] <SpamapS> nOStahl: if you just want basic virtualization.. both of them are pretty much overkill. |
|
[23:15] <nOStahl> so openstack eh? |
|
[23:16] <SpamapS> nOStahl: libvirt will probably do most of what you want for what you just described. |
|
[23:16] <SpamapS> nOStahl: clouds are more about being able to scale much much bigger. How many physical CPU's and GB of RAM do you want to support long term? |
|
[23:17] <nOStahl> aye thats what I was thinking |
|
[23:17] <nOStahl> can add nodes as needed |
|
[23:17] <nOStahl> and for redundancy |
|
[23:17] <nOStahl> thats a question I was having btw |
|
[23:18] <nOStahl> if I launch an instance and its running my webserver |
|
[23:18] <nOStahl> and i have two nodes and one dies |
|
[23:18] <SpamapS> nOStahl: openstack has no real notion of redundancy from what I've seen. Eucalyptus just now added HA in their 3.0 release. |
|
[23:18] <nOStahl> the instance still runs? |
|
[23:18] <SpamapS> nOStahl: hahaha |
|
[23:18] <SpamapS> hahahahhaha |
|
[23:18] <SpamapS> sorry I'm not laughing at you ;) |
|
[23:18] <SpamapS> nOStahl: that sounds pretty magical though, doesn't it? |
|
[23:18] <nOStahl> yes it does |
|
[23:18] <SpamapS> magic is bad mmkay :) |
|
[23:18] <nOStahl> heh |
|
[23:19] <nOStahl> is what I'm talking ha? |
|
[23:19] <SpamapS> nOStahl: there is no magic bullet for node HA. You're going to lose something if you make things available. See google searches for "CAP Theorem" |
|
[23:21] <nOStahl> so run me through how it works |
|
[23:21] <nOStahl> i launch an instance and it pushes the image over the network to one node? or many nodes? |
|
[23:21] <SpamapS> nOStahl: one |
|
[23:22] <nOStahl> ah I was thinking that it kind of mirrored it some how |
|
[23:22] <SpamapS> nOStahl: can you imagine how slow it would be if it mirrored all RAM access? |
|
[23:23] <SpamapS> nOStahl: there are "live migration" capabilities, where you can move nodes from one machine to another for expected downtime. |
|
[23:24] <SpamapS> nOStahl: and if you are using nova-volume with iSCSI, your disk will be stored somewhere else, so upon dying, nova can re-start your instance somewhere else. |
|
[23:24] <nOStahl> that must be what I was thinking of |
|
[23:24] <SpamapS> nOStahl: for the most part, you need to think about failover and recovery at the OS level, not the cloud provider level. |
|
[23:25] <SpamapS> nOStahl: DRBD for your data works |
|
[23:28] <nOStahl> getting more complicated lol |
|
[23:33] <SpamapS> nOStahl: scale out is complicated. :-P |
|
[23:38] <nOStahl> so ubuntu 11.10 then? |
|
|