<html><head>
<meta http-equiv="Content-Type" content="text/html; charset=Windows-1252"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; color: rgb(0, 0, 0); font-size: 14px; font-family: Calibri, sans-serif; "><div>Along this type of line, I know we've talked about it before.</div><div><br></div><div>But is cells the right way that we want to go? Not that it isn't, but possibly at the summit we can talk about it in detail before pushing it into trunk.</div><div><br></div><div>I still really like the idea of making nova-compute nodes more 'dumb', then having a 'global' scheduler service which controls allocates to those compute nodes. Something like the current nova-scheduler but not hooked into the MQ. U could almost think of this 'global' scheduler service hooking into all 'clusters' message queues and it would make the decisions about what nodes get scheduled where. What or how the 'global' scheduler works could be sharded, it could be hierarchal, but it wouldn't be tied to nova tightly. Which is what cells/zones (the previous attempt)/… seem to be. Perhaps a generic scheduling 'entity' where one of the resources is compute nodes would be the best thing overall to solve this (since volumes also need scheduling, networks probably could also and so on). </div><div><br></div><div>I just think in the long term that might be the best approach, but other thoughts are welcome…</div><div><br></div><span id="OLK_SRC_BODY_SECTION"><div style="font-family:Calibri; font-size:11pt; text-align:left; color:black; BORDER-BOTTOM: medium none; BORDER-LEFT: medium none; PADDING-BOTTOM: 0in; PADDING-LEFT: 0in; PADDING-RIGHT: 0in; BORDER-TOP: #b5c4df 1pt solid; BORDER-RIGHT: medium none; PADDING-TOP: 3pt"><span style="font-weight:bold">From: </span> Chris Behrens <<a href="mailto:cbehrens@codestud.com">cbehrens@codestud.com</a>><br><span style="font-weight:bold">Date: </span> Wednesday, October 3, 2012 4:57 AM<br><span style="font-weight:bold">To: </span> Sam Morrison <<a href="mailto:sorrison@gmail.com">sorrison@gmail.com</a>><br><span style="font-weight:bold">Cc: </span> "<a href="mailto:openstack@lists.launchpad.net">openstack@lists.launchpad.net</a>" <<a href="mailto:openstack@lists.launchpad.net">openstack@lists.launchpad.net</a>>, Chris Behrens <<a href="mailto:cbehrens@codestud.com">cbehrens@codestud.com</a>><br><span style="font-weight:bold">Subject: </span> Re: [Openstack] Cells Status<br></div><div><br></div><div><div style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; ">
Ok. This took a lot longer to resolve than I expected, but here we go:
<div><br></div><div><a href="https://github.com/comstud/nova/tree/cells_service">https://github.com/comstud/nova/tree/cells_service</a></div><div><br></div><div>This is rebased against trunk and contains a bunch of new things since the last branch:</div><div><br></div><div>Random fixes for things that trunk broke with cells (deleting instances for one)</div><div>RPC versioning (Thanks to Brian Elliott!)</div><div>Split Replies and Bandwidth Updates into their own queues to better deal with them</div><div>A number of admin API extensions modified to support cells (Thanks to Dragon, Alex Meade, Brian Lamar, Matt Sherborne, et al)</div><div>Snapshots/backups query glance in API cell (Thanks to Iccha)</div><div>Handle quotas in API cell (Thanks to Johannes for fixes!)</div><div><br></div><div>Things are rapidly getting more kludgy because of changes in trunk that don't have any consideration for cells (because cells is not in trunk!). I'm hoping we can get this into an acceptable shape such that we can get it merged ASAP.</div><div><br></div><div>- Chris</div><div><br></div><div><div><br><div><div>On Oct 2, 2012, at 10:13 PM, Sam Morrison <<a href="mailto:sorrison@gmail.com">sorrison@gmail.com</a>> wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite">OK great, will be good to get this into master. I have some stuff relating to key pairs, security groups that I'd like to contribute.<br><br>
Also we are looking at the ability for you to specify the cell when booting an instance.<br><br>
Cheers,<br>
Sam<br><br><br>
On 02/10/2012, at 1:06 PM, Chris Behrens <<a href="mailto:cbehrens@codestud.com">cbehrens@codestud.com</a>> wrote:<br><br><blockquote type="cite">Yup, it's done. I just have to deal with some conflicts with our internal branch and my public one..<br><br><br>
On Oct 1, 2012, at 7:47 PM, Sam Morrison <<a href="mailto:sorrison@gmail.com">sorrison@gmail.com</a>> wrote:<br><br><blockquote type="cite">On 02/10/2012, at 12:33 PM, Chris Behrens <<a href="mailto:cbehrens@codestud.com">cbehrens@codestud.com</a>> wrote:<br><br><blockquote type="cite">Thanks, Tom! I have changes to push up that add rpc versioning, etc. Maybe I can get those up tomorrow.<br></blockquote><br>
Great! I was going to start looking into it but will hold off if<br>
you've already done it.<br><br>
Cheers,<br>
Sam<br></blockquote><br></blockquote><br></blockquote></div><br></div></div></div></div></span></body></html>