=== 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 | ||
thomi | is it still the case in precise that I can't use juju to deploy to a rackspace cloud instance? | 21:23 |
---|---|---|
SpamapS | thomi: There's no RS cloud provider been written, no. | 23:09 |
thomi | SpamapS: ok, thanks. | 23:09 |
nOStahl | hi guys | 23:11 |
nOStahl | i got my hard drives in today | 23:11 |
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? | 23:12 |
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. | 23:13 |
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 | 23:14 |
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? | 23:15 |
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? | 23:16 |
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 | 23:17 |
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 | 23:18 |
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" | 23:19 |
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 | 23:21 |
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? | 23:22 |
SpamapS | nOStahl: there are "live migration" capabilities, where you can move nodes from one machine to another for expected downtime. | 23:23 |
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. | 23:24 |
SpamapS | nOStahl: DRBD for your data works | 23:25 |
nOStahl | getting more complicated lol | 23:28 |
SpamapS | nOStahl: scale out is complicated. :-P | 23:33 |
nOStahl | so ubuntu 11.10 then? | 23:38 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!