<div dir="ltr"><div>Hi,<br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Oct 15, 2021 at 7:10 PM James Penick <<a href="mailto:jpenick@gmail.com">jpenick@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr">This is something we've talked about doing at Yahoo some day. There are three separate problems to solve:<div><br></div><div>1. Diskless booting the compute node off the network. Mechanically this is possible via a number of approaches. You'd have a ramdisk with the necessary components baked in, so once the ramdisk loaded you'd be in the OS. I'm not sure if this can be fully accomplished via Ironic as yet. I'd need to ask an Ironic expert to weigh in.</div></div></blockquote><div><br></div><div><a href="https://docs.openstack.org/ironic/latest/admin/ramdisk-boot.html">https://docs.openstack.org/ironic/latest/admin/ramdisk-boot.html</a></div><div><br></div><div>Dmitry<br></div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div>2. Configuration of the compute node. Either a CI job which is aware of the compute node coming up and pushing configuration via something like Ansible, or perhaps using cloud-init with the necessary pieces loaded into a config-drive image which is provided as a part of the boot process. If we can have Ironic manage diskless booting systems then this would be a solved problem with user data.</div><div>3. VM storage could either be "local" via a large ramdisk partition (assuming you have a sufficient quantity of ram in your compute nodes), an NFS share which is mounted to the compute node, or volume backed instances.</div><div><br></div><div>We were investigating this earlier this year and got stuck on the third problem. Local storage via ramdisk isn't really an option for us, since we already pack our compute nodes with a lot of ram, and we need that memory for the instances. NFS has issues with security, since we don't want one giant volume exported to all compute nodes due to security concerns, and a per-compute node export would need to be orchestrated. Volume backed instances seemed ideal, however we ran into some issues there, which are partially related to the block storage product we use. I'm hopeful we'll get back to this next year, a class of instance flavors booted on diskless compute nodes would allow us to offer even more cost-effective options for our customers.</div><div><br></div><div>-James</div><div><br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Oct 15, 2021 at 3:54 AM A Monster <<a href="mailto:amonster369@gmail.com" target="_blank">amonster369@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr">In<span style="color:rgb(35,38,41);font-family:-apple-system,BlinkMacSystemFont,"Segoe UI","Liberation Sans",sans-serif;font-size:15px"> Openstack, is it possible to create compute nodes with no hard drives and use PXE in order to boot the host's system and therefore launch instances with no local drive which is needed to boot the VM's image.</span><p style="margin:0px;padding:0px;border:0px none;font-variant-numeric:inherit;font-variant-east-asian:inherit;font-stretch:inherit;line-height:inherit;font-family:-apple-system,BlinkMacSystemFont,"Segoe UI","Liberation Sans",sans-serif;font-size:15px;vertical-align:baseline;box-sizing:inherit;clear:both;color:rgb(35,38,41)">If not, what's the minimum storage needed to be given to hosts in order to get a fully functional system.</p></div>
</blockquote></div>
</blockquote></div><br clear="all"><br>-- <br><div dir="ltr" class="gmail_signature"><div dir="ltr">Red Hat GmbH, <a href="https://de.redhat.com/" target="_blank">https://de.redhat.com/</a> , Registered seat: Grasbrunn, <br>Commercial register: Amtsgericht Muenchen, HRB 153243,<br>Managing Directors: Charles Cachera, Brian Klemm, Laurie Krebs, Michael O'Neill <br></div></div></div>