[openstack-dev] [Glance] image.exists events
Hello, I recently created a script to generate image.exists events similar to what is described here: https://blueprints.launchpad.net/glance/+spec/glance-exists-notification I am wondering if it's a good idea to finish the implementation described in that spec ? image.exists events are useful especially for resource accounting/billing (what images were active for a tenant 7 months ago for example). (see nova instance.exists events or cinder, neutron and probably other projects) -- Cristian Tomoiaga __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
[openstack-dev] [nova] Nova and LVM thin support
Hello everyone, Before going any further with my implementation I would like to ask the community about the LVM thin support in Nova (not Cinder). The current implementation of the LVM backend does not support thin LVs. Does anyone believe it is a good idea to add support for this in Nova ? (I plan on adding support for my implementation anyway). I would also like to know where Red Hat stands on this, since they are primarily working on LVM. I've seen that LVM thin would be supported in RHEL 7 (?) so we may consider the thin target stable enough for production in Juno (cinder already has support for this since last year). I know there was ongoing work to bring a common storage library implementation to oslo or nova directly (Cinder's Brick library) but I heard nothing new for some time now. Maybe John Griffith has some thoughts on this. The reasons why support for LVM thin would be a nice addition should be well known especially to people working with LVM. Another question is related to how Nova treats snapshots when LVM is used as a backend (I hope I didn't miss anything in the code): Right now if we can't do a live snapshot, the instance state (memory) is being saved (libvirt virDomainManagedSave) and qemu-img is used to backup the instance disk(s). After that we resume the instance. Can we insert code to snapshot the instance disk so we only keep the instance offline just for a memory dump and copy the disk content from the snapshot created ? -- Regards, Cristian Tomoiaga ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] Debian Jessie freeze date announced: 5th of November 2014
Hello Thomas, I am sorry to send a reply a little late on this. I plan on working with Debian for my Openstack setups (now I'm on a rhel based setup) and I would really like the latest OpenStack release available. I was initially planning to setup my own mirrors since I always seem to need features from the next OpenStack release. For example Grizzly for me looks too old and some features that were supposed to land on Havana are now scheduled for Icehouse. Given this, I would pretty much like to have J in Debian Jessie. I'm not sure how to approach this or if it's worth the effort on your part given the latest issues you submitted for Havana and since most likely some features in K will probably make me switch to separate mirrors anyway. However, taking into account the rapid development of OpenStack my guess is that the J release should land in Jessie if possible. I will also try to find some time and help out as much as I can with this. Let me know what you decide , probably after the summit. On Mon, Oct 14, 2013 at 7:34 AM, Thomas Goirand z...@debian.org wrote: Hi, The Debian release team has announced the release date for Jessie: https://lists.debian.org/debian-devel-announce/2013/10/msg4.html This means that for this release, we will *not* have any kind of sync with Ubuntu LTS (last time for Wheezy, we froze a few months after the 2012.04 LTS). I haven't made up my mind yet if we should release Debian Jessie with OpenStack Icehouse (to have the same release as the Ubuntu LTS), or with the J release. I'd be happy to gather comments and suggestions about it, and discuss about it at the HK summit. Cheers, Thomas Goirand (zigo) ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev -- Regards, Cristian Tomoiaga ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova] RFC: adding on_shared_storage field to instance
Hello Chris, Just a note regarding this. I was thinking on using local plus shared storage for an instance ( eg. root disk local and another disk as a cinder volume ). If I understand this correctly, flagging the instance as having local storage may not be such a good idea in this particular case right ? Maybe root_on_local ? Regards -- Regards, Cristian Tomoiaga ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
[openstack-dev] [openstack][neutron] Handling issues unlikely to happen with the current code
Hello, I am wondering if we should handle potential issues (unlikely to happen with the current code) now or wait to see what happens with the code in the future ? What I am referring to can be seen for example in neutron/db/db_base_plugin_v2.py in _recycle_ip if we try to recycle the same IP twice. Looking at where and how the _recycle_ip function is called, it's unlikely that the same IP will be recycled twice, but if that happens, two allocation pools will be created containing the same IP address leading to all sorts of issues afterwards. There is no check to see if the IP already exists as a single entry in the allocation pools table. An extra db query should be created to avoid this. The IP handling code will change in the future but I am guessing the same function will continue to exist. Waiting for some feedback on this, thank you! -- Regards, Cristian Tomoiaga ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Neutron][Climate] bp:configurable-ip-allocation
Hello Nikolay, Looking at this bp, it seems it has been targeted for icehouse-1 :( I was waiting for this too (for some time now). Mark I may be able to help if needed (will this use the same logic as the abandoned code ?). I am working on something similar to floating IPs but for normal IPs. I need to be able to allow project owners to reserve specific IPs and allocate them to VMs as needed (targeted at flat, provider networks where project owners need to keep IP addresses) -- Regards, Cristian Tomoiaga ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
[openstack-dev] [openstack][neutron] Reserved fixed IPs
Hello everyone, I am working on implementing fixed IP reservation for tenants. My goal is to be able to reserve fixed IPs for a tenant and avoid as much as possible the ephemeral state of an IP. A basic workflow would be like this: Tenant or admin reserves one or more fixed IPs. He will than be able to use one or more of those reserved IPs on his instances (assign them to ports, support multiple IPs per port). If no/not enough fixed IPs are reserved, use the current IPAM implementation otherwise allow the tenant to select from his reserved IPs and then go through the current IPAM. I am using fixed routable and non-routable IPs for public and private networks (provider network , no NAT and no tagging). I will also use floating IPs for LB, DNS a.s.o. I have a few questions regarding the development of this since the documentation is still being worked on and I have to dig through the code a lot to understand a few things: 1. nova reserve-fixed-ip, this belongs to nova-network now obsolete right ? 2. I though of creating a new model (mainly a db table) to hold the IPs and the tenant IDs in order to keep the association. I've done this for the openvswitch plugin in ovs_models_v2 by adding a new model. I can probably do this globally in /db directly right (especially if I plan on supporting multiple plugins) ? 3. I was planning on adding to Neutron the api calls nova has for fixed IPs (ex: fixed-ip-get, reserve, unreserve) Does this seem right ? I am asking because I believe there is some work towards a new IPAM implementation and I would like to get some thoughts. I am also asking because to me it seems a little bit confusing that nova can also manage IPs and I am not sure if/what functions are obsolete there. 4. This should go as an extension first (as far as I understand for the docs). Add the extension to extend the Neutron API and modify the current IPAM right ? -- Regards, Cristian Tomoiaga ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev