Conversation with #opennebula at Thu 16 May 2013 10:39:17 AM CEST on firstname.lastname@example.org (irc)
(05:05:16 PM) jmelis: Ok, so let's begin then
(05:05:25 PM) jmelis: this is going to be a more open discussion
(05:05:43 PM) jmelis: no need for another channel like in previous opennebula IRC meetings
(05:05:57 PM) jmelis: we're going to structure the meeting like this
(05:06:20 PM) jmelis: 1) we will review and explain the features that we've selected for OpenNebula 4.2
(05:06:49 PM) MLodi [~email@example.com] entered the room.
(05:06:56 PM) MarcelloLodi left the room.
(05:07:20 PM) jmelis: 2) we will discuss the requests provided by the community and after we've collected feedback, we will decide at a later time if they can make it to the release
(05:07:42 PM) jmelis: In any case, feel free to interrupt and ask any questions whenever you see fit
(05:08:02 PM) Ruetobas left the room (quit: ).
(05:08:05 PM) jmelis: so, let's start by the features we have selected
(05:08:18 PM) jmelis: you can see them here http://dev.opennebula.org/projects/opennebula/issues?query_id=36
(05:08:24 PM) jmelis: divided into categories
(05:09:01 PM) jmelis: carlosms: can you explain a bit the work that will be done regarding AppFlow?
(05:10:02 PM) carlosms: ok... so the main goal for 4.2 will be to integrate the apps into the opennebula distribution
(05:10:14 PM) carlosms: and regarding the appflow component specifically
(05:10:32 PM) carlosms: we are going to focus on providing elasticity
(05:10:56 PM) carlosms: to scale up/down the services based on monitoring information gathered by opennebula
(05:11:34 PM) carlosms: we will also try to develop a way to allow VM guests to contact opennebula
(05:11:56 PM) rolnas: metadata-server
(05:12:01 PM) carlosms: and add custom monitoring information from any service running on the VM
(05:12:21 PM) carlosms: that's it, rolnas
(05:12:23 PM) carlosms: :)
(05:12:41 PM) Nostalgeek: kinda goes along # 1768
(05:13:02 PM) stefunel: and how will this monitoring information be gathered? with an agent?
(05:13:28 PM) carlosms: http://dev.opennebula.org/issues/2054
(05:13:52 PM) jfontan: I think it is a bit different from 1768
(05:14:00 PM) carlosms: we'll do a small REST server that will be agnostic to anything running inside the VM
(05:14:21 PM) carlosms: to allow you to use curl, ruby, whatever you have available....
(05:14:53 PM) Nostalgeek: ok, the metadata-server and cloudinit are similar to me but that's fine :)
(05:14:57 PM) ricduarte: will you make it compatible with AWS?
(05:14:58 PM) stefunel: it's a really nice idea, but I think making this really really secure will be problematic
(05:15:20 PM) ricduarte: actually the metadata server is not the same as cloud init
(05:15:35 PM) Ruetobas [Ruetobas@122-116-54-163.HINET-IP.hinet.net] entered the room.
(05:15:53 PM) ricduarte: metadata-server supports all integration scripts compatible with EC2
(05:15:57 PM) carlosms: stefunel: yeah, that's a concern. We'll see how we can generate individual tokens and such
(05:15:58 PM) ricduarte: not just cloudinit
(05:16:26 PM) ricduarte: carlos: any chance of making opennebulaapps compatible with EC2 apis?
(05:16:26 PM) Nostalgeek: but cloudinit does contact the metadata server
(05:16:38 PM) dmolina: this metadata server can be used to push and not only to pull info
(05:16:39 PM) rolnas: tokens would be nice for secure shared storage mounting
(05:17:08 PM) carlosms: ricduarte: that's not really on our short term roadmap
(05:17:22 PM) dmolina: so a vm can notify when its ready or modify elasticity rules
(05:18:27 PM) rsmontero: ricduarte: it'll be functional compatible as you can get the context inforation, i.e. the metadata-tags. However, it'll not be 100% compatible because of the 169.254.169.254 IP
(05:18:48 PM) ricduarte: that does not seem a problem
(05:18:55 PM) ricduarte: you can easilly nat it to the one controller
(05:19:47 PM) rsmontero: I see, show we can just add also the EC2 route to the server .... latest/meta-data/instance-id
(05:20:27 PM) jmelis: s/show/so/ I guess :)
(05:20:34 PM) rsmontero: :P
(05:20:53 PM) Nostalgeek: i like that, i think we should try to stick to 169.254.169.254
(05:21:09 PM) ricduarte: there is also a dns name
(05:21:14 PM) ricduarte: but not every implementation uses it
(05:21:20 PM) rsmontero: What about the auth mechanism as in EC2 you do not need to send any token from the VM
(05:21:27 PM) rolnas: what about ec2 keypairs for ldap/external auth users, for the moment we have to create additional accounts to use cloud apis (ec2/occi)
(05:21:41 PM) ricduarte: in my installation
(05:21:51 PM) jmelis: Nostalgeek: The CloudInit stuff has been planned for ONE 4.2, but the approach we're going to follow is to make CloudInit work with the OpenNebula Context CDROM
(05:22:06 PM) jfontan: well, thinking again nostalgeek's idea of one metadata server with pull (for configuration) and push (for monitoring) is a nice idea
(05:22:26 PM) jmelis: apparently in CloudInit you can define data sources, and this can be either 169.254... or a path, where the data should be pickled
(05:22:42 PM) jfontan: the same access control that is already there to get information can be used to push information
(05:23:08 PM) jmelis: jfontan: yes it is, but CloudInit already supports static files as data sources, which is the way OpenNebula already works
(05:23:09 PM) Nostalgeek: jmelis, OK if we can do this without forking Cloudinit I don't see why not.. i just hope we can ditch the context scripts in favor of a more standard Cloudinit approach
(05:23:12 PM) ricduarte: did you had a look to http://docs.aws.amazon.com/AutoScaling/latest/APIReference/Welcome.html ?
(05:23:24 PM) ricduarte: why not mimic EC2 approach?
(05:23:24 PM) Nostalgeek: my secret goal is to run un-modified EC2 image on my OpenNebula clouad :)
(05:23:30 PM) jmelis: Nostalgeek: that's are long-term approach, yes
(05:23:36 PM) ricduarte: nostalgeek: I can run unmodified EC on opennebula
(05:23:45 PM) jfontan: jmelis, I agree, but I'm just thinking on reusing things that are already done for the monitoring server
(05:23:51 PM) ricduarte: you just have to install the metadata-server, and have dhcp available
(05:23:57 PM) Nostalgeek: ricduarte, you can, but have to turn off cloudinit or else run into lengthly boot time because of it
(05:24:02 PM) ricduarte: nop
(05:24:07 PM) ricduarte: works like EC2
(05:24:11 PM) ricduarte: boots in seconds
(05:24:17 PM) Nostalgeek: ricduarte, or use metadata-server :) but I dont
(05:24:29 PM) rolnas: I already got ubuntu cloud image to run on our testing opennebula 4.0.0 with metadata-server from ricduarte
(05:24:32 PM) ricduarte: metadata + 169.254.169.254 NAT + dhcp => unchanged
(05:24:42 PM) ricduarte: you need a serial console tho
(05:24:54 PM) Nostalgeek: ok
(05:25:01 PM) ricduarte: and to change the datastore driver to increase the size of the images
(05:25:05 PM) ricduarte: beucase they are generally small
(05:25:17 PM) Nostalgeek: yes, ran into this too
(05:25:31 PM) ricduarte: with cloudinit you just need to increase the size of the qcow
(05:25:41 PM) ricduarte: actually this relates to the cloudinig feature on 4.2
(05:25:41 PM) Nostalgeek: So I guess we really need to look at retaining as much compatibility with the EC2 cloudinit / metadata approach
(05:26:08 PM) ricduarte: ec2 already has apis for monitoring and auto scalling
(05:26:14 PM) jmelis: ricduarte: the image resize is also a planned feature
(05:26:20 PM) rolnas: in parallel we need to make some way to auth ec2 for external auth users
(05:26:25 PM) ricduarte: rolnas
(05:27:14 PM) ricduarte: did you say #1391 ?
(05:27:17 PM) ricduarte: *see
(05:27:35 PM) ricduarte: http://dev.opennebula.org/issues/1391
(05:27:58 PM) rolnas: yes, I see
(05:28:12 PM) ricduarte: we can't make LDAP auth dependant on a MD5 or plain text, because it will not support activedirectory
(05:28:46 PM) Nostalgeek: jfontan, regarding using metadata server for push I'm a not sure. To me cloudinit and metadata server are truly one, and I fear we get a bit for off the cloudinit goal if we implement push from the VM. Perhpas we should see if there is anything in Cloudinit that allows VM to push to the metata
(05:29:12 PM) rsmontero: so, it seems that EC2 compatibility is important for some of you. Could you elaborate that a bit: Are you using econe server for your users? Is that better than the new sunstone views? In general, would you like to see more EC2 API calls implemented?
(05:29:18 PM) rolnas: ricduarte: but we can add additional attributes to ldap user entry for ec2/occi auth
(05:29:23 PM) ricduarte: Nostalgeek: you can have a new namespace
(05:29:38 PM) ricduarte: rolnas: people don't like to add stuff to AD
(05:29:54 PM) ricduarte: I'm currently doing as #1391 and works great
(05:30:03 PM) ricduarte: and there is nothing changed @ LDAP
(05:30:07 PM) rolnas: ricduarte: I mean to add attributes for opennebula user
(05:30:12 PM) rolnas: in opennebula DB
(05:30:17 PM) Nostalgeek: rsmontero, to me it's only about running unmodified EC2 images. I'm happy with the API and Sunstone
(05:30:32 PM) jmelis: Nostalgeek: good point
(05:30:44 PM) ricduarte: why do you want to push with cloud-init?
(05:31:03 PM) rolnas: ec2 is important for compatibility with defacto standard of clouds
(05:31:13 PM) ricduarte: but what do you want to upload?
(05:31:17 PM) Nostalgeek: ricduarte, not sure who you're asking? i don't need push with cloud-init
(05:31:28 PM) jmelis: Nostalgeek: but why do you care if CloudInit uses that metada-server for the data source for contextualization or the OpenNebula Context CDROM?
(05:31:39 PM) Nostalgeek: i think that's something the NebulaApp users need
(05:32:02 PM) ricduarte: @jmelis, using cloud-init makes unchanged images run
(05:32:19 PM) Nostalgeek: jmelis, because the context CDROM are Opennebula only
(05:32:26 PM) ricduarte: also, as far as I know currenyl appflow requires changes to the image as well, for the "chef" glue
(05:32:48 PM) ricduarte: with cloud-init + user-data you don't need to change anything, because it supports chef, puppet, etc out of the box
(05:33:02 PM) jmelis: Nostalgeek, ricduarte: yes, but that's precisely the effor planned http://dev.opennebula.org/issues/1768
(05:33:08 PM) rolnas: Random password generation per instance (for e.g. for VNC/PASSWORD or VM root/user passwords) would be nice and I tried already to find where to put this code
(05:33:16 PM) jmelis: we intend to patch CloudInit *upstream* so it support OpenNebula Context CDROM
(05:33:30 PM) ricduarte: rolnas: you can add it to the metadata-server
(05:33:38 PM) Nostalgeek: I think we're going the wrong way with the context scripts right now, we should look at a more standard way of doing the contextualization stuff, thus my #1768 for (ultimately) replacing context by cloudinit
(05:33:38 PM) ricduarte: now with 4.0 you can write to the vm template
(05:34:03 PM) jmelis: Nostalgeek: CloudInit and OpenNebula COntext CDROM are not incompatible IMO
(05:34:06 PM) ricduarte: actually I like context scripts alot
(05:34:17 PM) ricduarte: and they are not incompatible with cloudinit
(05:34:32 PM) Nostalgeek: perhaps we can do both, then :)
(05:34:36 PM) Nostalgeek: i dont use context scripts here
(05:34:51 PM) Nostalgeek: but looking at using cloudinit
(05:35:08 PM) ricduarte: Nostalgeek: they are quite valuable to contextualize windows
(05:35:15 PM) ricduarte: and to avoid DHCP
(05:35:25 PM) ricduarte: you can't do IP configuration with cloud-init
(05:35:51 PM) ricduarte: (you can, but not with the EC2 module)
(05:35:56 PM) Nostalgeek: humm ok. good point then.. jmelis are any of you guys on the cloudinit mailing list or project dev ?
(05:36:09 PM) Nostalgeek: perhaps these are questions we could ask them (is there a way to configure IP addresses, etc.)
(05:36:27 PM) ricduarte: Nostalgeek: you can do whatever you like, they are python scripts
(05:36:31 PM) rolnas: I have working prototype of virtual router in bare metal, but I would like to make it accessible with API to be more dynamic, I like how it works in openstack :-)
(05:36:34 PM) jfontan: this cloud-init version configures the network with opennebula context https://code.launchpad.net/~vlastimil-holer/cloud-init/opennebula
(05:36:46 PM) ricduarte: actually, I guess the openstack cloudrive module shoud support it
(05:36:56 PM) jmelis: Nostalgeek: jfontan is the one looking at that
(05:37:08 PM) ricduarte: maybe you could also put a cloudrive manifest on the context drive
(05:37:34 PM) jmelis: rolnas: the metadata-server will be precisely for those things
(05:37:43 PM) Nostalgeek: ok, well, I like the cloudinit dicussion we're having, perhaps we could continue this a bit later
(05:37:54 PM) jmelis: enabling communication between OpenNebula and the VMs
(05:37:55 PM) rolnas: I would like also make bare metal provisioning in opennebula, like it works in current openstack
(05:38:22 PM) jmelis: and having a Virtual Router with communication with OpenNebula would be a wonderful feature
(05:38:22 PM) ricduarte: maybe we could have a look at puppet and razor for that
(05:38:46 PM) jmelis: rolnas: I'm also very interested in the bare metal feature
(05:38:49 PM) tinova: rolnas: what specific functionality for bare metal you have in mind?
(05:38:58 PM) jmelis: could you give it a thought and create a request?
(05:39:08 PM) jmelis: explaining how you think the integration should be done, etc?
(05:39:15 PM) rolnas: idea is to manage everything in DC with opennebula
(05:39:25 PM) ricduarte: cloudstack also supports bare metal, maybe you can have a look
(05:39:26 PM) rolnas: from bare metal to VM
(05:39:57 PM) ricduarte: rolnas: did you look at kanopya?
(05:41:06 PM) jmelis: so, let me wrap up some ideas here: we have agreed that CloudInit + OpenNebula Context CDROM is a good feature to have
(05:41:22 PM) jmelis: the Appflow server provides a metadata-server
(05:41:28 PM) ricduarte: @jmelis, I would add support for clouddrive; it's easilly implementable
(05:41:30 PM) jmelis: and it's a good idea to route it using nat
(05:41:42 PM) Nostalgeek: rolnas, regarding password generation, one thing I am looking at is clearing that attribute from OpenNebula once the VM has been deployed. This is something possible now with the one.vm.update call
(05:42:22 PM) jmelis: so, moving on
(05:42:30 PM) jmelis: we've already discussed the AppFlow component
(05:42:35 PM) jmelis: the Context component too
(05:42:52 PM) jmelis: there are no big ones in the CLI section
(05:42:59 PM) jmelis: and that brings us to the Core
(05:43:05 PM) jmelis: there are two big ones here, at least
(05:43:10 PM) jmelis: #1613
(05:43:25 PM) jmelis: http://dev.opennebula.org/issues/1613
(05:43:34 PM) jmelis: http://dev.opennebula.org/issues/2032
(05:44:12 PM) jmelis: (and http://dev.opennebula.org/issues/2009 but this one will probably make it for 4.0.1 :D )
(05:44:28 PM) Nostalgeek: regarding # 1613
(05:44:39 PM) Nostalgeek: could we use this to keep qcow2 image size in sync ?
(05:45:10 PM) Nostalgeek: (I don't use qcow2 here but noticed a discussion linked with the image resize issue)
(05:45:16 PM) jmelis: yes, that would be the same campaign
(05:45:20 PM) jmelis: take a look at #1662
(05:45:48 PM) Nostalgeek: yes, #1770 was the one i was talking about
(05:45:54 PM) rolnas: also about qcow2 image performance optimizations (like metadata preallocation)
(05:46:19 PM) jmelis: Nostalgeek: yes, they're all related and it will be supported
(05:46:37 PM) rolnas: I have some changes in remotes/tm/qcow2 and found that QEMU_IMG is not really used in several places
(05:46:40 PM) Nostalgeek: jmelis, thanks
(05:48:37 PM) jmelis: dmolina: any sunstone goodness for 4.2?
(05:48:47 PM) rolnas: better stability
(05:48:50 PM) rolnas: :-)
(05:49:02 PM) ricduarte: I miss the self service portal
(05:49:18 PM) dmolina: yeah in this release we will try to focus on fixing most of the issues you have openned and we have found
(05:49:22 PM) rolnas: for the moment sunstone 4.0 looks rough to sunstone 3.8
(05:49:58 PM) dmolina: we will implement some missing features such as default quotas, enforce parameter in live migration actions, deatacable vnc
(05:50:05 PM) rolnas: idea of self-service portal is good also, but implementation in 3.8 was very limited
(05:50:15 PM) dmolina: ricduarte: currently you can get it with the new sunstone views
(05:50:22 PM) ricduarte: actually no
(05:50:26 PM) dmolina: you can enable disable tabs/actions/columns
(05:50:38 PM) dmolina: what are you missing?
(05:50:39 PM) ricduarte: what I miss most is the ability to have multiple instance types for the same image
(05:50:41 PM) ricduarte: without creating templates
(05:50:54 PM) ricduarte: like openstack, ec2, cloudstack
(05:50:55 PM) rolnas: like in ec2
(05:51:09 PM) ricduarte: so much i was forced to do my own "self-service" portal
(05:51:14 PM) ricduarte: :[
(05:51:23 PM) dmolina: with the old self service you have to create a new file for each instance type
(05:51:30 PM) dmolina: now you can use the opennebula template
(05:51:36 PM) dmolina: permissions and such
(05:51:36 PM) ricduarte: @dmolina, i'm doing in the following wat
(05:51:44 PM) ricduarte: have a generic template for windows and linux instances
(05:51:47 PM) rolnas: template != instance type
(05:51:52 PM) Nostalgeek: ricduarte, you can specify the DISK when instantiating a template
(05:51:56 PM) ricduarte: users can select image + instance type
(05:52:20 PM) ricduarte: instance types defined within one documents
(05:52:31 PM) dmolina: that functionality will be included for 4.2
(05:52:42 PM) dmolina: mergin info when instantiating a new template
(05:52:42 PM) ricduarte: very nice
(05:52:48 PM) dmolina: currently you can do it through the cli
(05:52:53 PM) rolnas: when 3.8.4 will be
(05:52:54 PM) rolnas: ?
(05:53:09 PM) dmolina: so you can define diferrent templates just specifying cpu and memory
(05:53:19 PM) dmolina: and when you instantiate it you can select the image
(05:53:21 PM) ricduarte: any chance of adding #2048 to core?
(05:53:46 PM) Nostalgeek: here I have something very similar to Amazon Small, Medium, Large etc... they are all template in OpenNebula. And when instantiating new VM I specify which DISK i want attached, not sure thats what you mean ricduarte
(05:54:09 PM) ricduarte: something like that
(05:54:26 PM) ricduarte: I also added a image catalog and ability to select keypair
(05:54:30 PM) ricduarte: also elastic ip btw
(05:54:39 PM) ricduarte: something that I also find missing @sunstone
(05:55:35 PM) ricduarte: self-service should really match what openstack does
(05:55:42 PM) dmolina: rolnas: most of the problems you resolved will be fixed for 4.0.1, did you find anything else?
(05:56:02 PM) rolnas: what about 3.8.4 ?
(05:56:04 PM) dmolina: s/resolved/reported
(05:56:06 PM) Nostalgeek: ricduarte, for elastic IP I'm looking at something like # 1818
(05:56:19 PM) jmelis: rolnas: next week
(05:56:28 PM) Nostalgeek: which is planned for 4.2
(05:56:30 PM) ricduarte: but it needs to run scripts
(05:57:07 PM) Nostalgeek: using reserve + attachnic you can do the EC2 elastic IP
(05:57:35 PM) ricduarte: but how do you run the script that really configures the ip?
(05:57:36 PM) ricduarte: hooks?
(05:57:44 PM) Nostalgeek: attachnic VMM
(05:57:52 PM) ricduarte: uhmm....
(05:58:05 PM) dmolina: http://opennebula.org/documentation:rel4.0:ec2qcg#enabling_elastic_ip_functionality
(05:58:26 PM) rolnas: sunstone needs ssh key management
(05:58:31 PM) rolnas: for users
(05:58:42 PM) rolnas: ssh public keys
(05:58:43 PM) ricduarte: @dmolina: problem is that this is only for econe; sunstone should support it; api should support it as well, etc
(05:58:57 PM) jmelis: rolnas: you can add your ssh key in your template
(05:59:02 PM) ricduarte: rolnas: you have the keypair already @USERS
(05:59:11 PM) Nostalgeek: dmolina, oh wow didn't notice that...
(05:59:23 PM) ricduarte: would be cool to make it available @sunstone and self-service
(05:59:30 PM) rolnas: I mean sunstone GUI to manage ssh public keys
(05:59:38 PM) dmolina: rolnas: there is also a ticket for including the ssh key when creating a new user http://dev.opennebula.org/issues/1830
(06:00:15 PM) Nostalgeek: ricduarte, the one.vm.attachnic API call will trigger the VMM attachnic action which is in charge of attaching the IP to the VM
(06:00:36 PM) dmolina: ricduarte: it would be great to include it in sunstone/oned, could you please open a ticket in the dev page
(06:00:49 PM) ricduarte: attachnic does not imply adding a nic?
(06:00:59 PM) Nostalgeek: ricduarte, kinda, for me 1 IP = 1 NIC
(06:00:59 PM) Nostalgeek: :)
(06:01:09 PM) ricduarte: but elastic does not work like that
(06:01:26 PM) ricduarte: generally, you have a load balancer/rotuer and you just NAT your elastic ip to your instance ip
(06:01:38 PM) Nostalgeek: I see what you mean
(06:01:39 PM) ricduarte: the instance is unchanged
(06:01:45 PM) jmelis: yeah, ricduarte is right, you need either openflow or nat rules
(06:01:49 PM) ricduarte: how about reserve IP + hooks?
(06:02:02 PM) rolnas: we want also have elastic ports, because external/public IPs are scarse
(06:02:17 PM) ricduarte: and how about elastic dns?
(06:02:44 PM) Nostalgeek: ricduarte, to me, the hooks play a major part in customizing opennebula, I'd love to see more hooks
(06:02:58 PM) rolnas: scarse -> scarce
(06:02:58 PM) jmelis: rolnas: there will be many possibilities opening once we have the metadata server
(06:03:06 PM) jmelis: to speak with a Virtual Router, etc...
(06:03:14 PM) Nostalgeek: ricduarte, the initial idea of reserve is to "hold" an IP for a give user, it doesn't necessarily trigger the binding with a VM
(06:03:31 PM) Nostalgeek: thus I am using attachnic
(06:03:31 PM) ricduarte: I see....
(06:03:55 PM) ricduarte: but maybe a binding would make sense
(06:03:56 PM) jmelis: let's move on, to sum up there's a lot of concern with elastic * features, and we need to open that road
(06:04:36 PM) Nostalgeek: ok
(06:04:43 PM) ricduarte: another thing missing is security groups
(06:04:44 PM) jmelis: and rolnas: it's important that you send the things you find "rough" as bug reports
(06:04:53 PM) dmolina left the room (quit: Ping timeout: 248 seconds).
(06:04:53 PM) jmelis: ricduarte: you have white_ports
(06:04:58 PM) rolnas: ok, ok
(06:05:14 PM) jmelis: which provides the same functionality
(06:05:30 PM) rolnas: when 4.2 is planned to release ?
(06:05:42 PM) jmelis: it's going to be a very short release
(06:05:45 PM) ricduarte: not the same
(06:05:48 PM) jmelis: July
(06:05:50 PM) jmelis: why not?
(06:05:51 PM) ricduarte: I was looking for soemthing like #1225
(06:05:54 PM) jmelis: ricduarte: why not?
(06:06:22 PM) jmelis: ricduarte: noted :)
(06:06:32 PM) jmelis: let's see if we can bump this one for 4.2
(06:06:36 PM) ricduarte: @jmelis: security groups kind of keep your rules in one place
(06:06:41 PM) ricduarte: and you can attach it to instances
(06:06:52 PM) ricduarte: and is available within ec2
(06:06:54 PM) ricduarte: etc
(06:07:23 PM) cdaffara left the room (quit: Ping timeout: 250 seconds).
(06:07:33 PM) Nostalgeek: just a quick thought about 1225 perhaps now that we have USER_TEMPLATE it would make sense for the firewall rules to be there
(06:07:36 PM) jmelis: keep in mind that we are not going to enhance the ec2 interoperability with ec2 for this release
(06:07:43 PM) Nostalgeek: and have an "UPDATE" hook that fires when the user updates the USER_TEMPLATE
(06:08:02 PM) jmelis: but I encourage you all to write requests with all the things you find missing
(06:08:32 PM) jmelis: Nostalgeek: noted
(06:08:49 PM) ricduarte: @jmelis, SG are not only related to EC2
(06:09:02 PM) ricduarte: @jmelis, it would be valuable for any api/cli etc
(06:09:15 PM) rolnas: do one have some failure mode for stuck ssh connections for monitoring of host/vm, because I have those situation when shared FS fails
(06:09:17 PM) rolnas: ?
(06:09:56 PM) jmelis: ricduarte: ok
(06:10:24 PM) jmelis: rolnas: none that I can think of at the moment
(06:10:45 PM) jmelis: Drivers category
(06:11:32 PM) jmelis: Open vSwtich improvement, LVM improvement, move xen's api to XL
(06:12:03 PM) jmelis: and I'll make a note for what rolnas just said, it could just fit in with # 2009
(06:13:33 PM) Nostalgeek: i really wish # 2019 could make it, to me hooks or a major part of what makes OpenNebula great
(06:13:40 PM) Nostalgeek: its "hackability" :)
(06:13:42 PM) jmelis: right so that sums up the Opennebula 4.2
(06:14:30 PM) jmelis: as you can see it's a short release, except for datastore monitoring (which is a huge issue) and the appflow component (which will be an even greater issue)
(06:15:03 PM) Nostalgeek: jmelis, I think you said the release is planned around july ?
(06:15:08 PM) jmelis: yes
(06:15:36 PM) Nostalgeek: and 4.4 for OpennebulaConf ? :)
(06:15:39 PM) rolnas: when doing datastore monitoring some of them can share the same space
(06:15:42 PM) jmelis: probably, yes :)
(06:15:55 PM) ricduarte: any chance of adding in #1727 ?
(06:16:02 PM) jmelis: right, so let me recap
(06:16:20 PM) Nostalgeek: ricduarte, i think 1727 is quite a bit of work
(06:16:35 PM) jmelis: these issues we've previously selected have been based on the feedback of the community etc
(06:17:19 PM) jmelis: but I would like to ask you if you could let me know of some very important issues to you for 4.2
(06:17:30 PM) jmelis: which brings us to Nostalgeek asking about 2019
(06:17:46 PM) jmelis: I'll make a note of this
(06:17:51 PM) Nostalgeek: jmelis, i haven't look at the hook path at all in the core
(06:17:59 PM) Nostalgeek: i don't know if its trivial or not..
(06:18:16 PM) rolnas: lets make 4.2 worth to migrate from 3.8 without big issues
(06:18:34 PM) jmelis: Nostalgeek: let's discuss that on the ticket
(06:18:38 PM) jmelis: rolnas: what issues did you have?
(06:18:42 PM) Nostalgeek: jmelis, thanks
(06:18:54 PM) rolnas: for the moment 4.0 doesn't look well for production
(06:19:15 PM) rolnas: it's my subjective look
(06:19:18 PM) rolnas: :-)
(06:19:31 PM) jmelis: can you list the main points you feel are not production ready?
(06:19:40 PM) rolnas: mostly sunstone
(06:19:46 PM) ricduarte: rsmontero: will the solution proposed @ last commect of #1727 be implemented?
(06:19:49 PM) rolnas: core - probably OK
(06:19:57 PM) jmelis: ok, that's what I thought
(06:20:15 PM) darkfader: how much did the monitor and vm state handling improve?
(06:20:27 PM) jmelis: rolnas: as you know sunstone has been written from the ground up, and as such, we will find rough edges here and there sometimes
(06:20:29 PM) rolnas: 3.8 self service portal probably will work with 4.0 also
(06:20:51 PM) rolnas: because of occi API
(06:21:20 PM) jmelis: rolnas: but 4.0.1 is coming up, with more sunstone fixes, and it's undergoing rapid development and maintenance, so you can rest assured that it will continue to improve very quicky
(06:21:38 PM) Nostalgeek: darkfader, I really (really) like the new VM POLLING from the host MONITOR. It means all your VMs running on the same host are monitored at the same time
(06:21:44 PM) darkfader: jmelis: actually, i think the worst bit about sunstone (for users) is the idea that anyone would prefer editing templates over giving them menus where they do stuff graphically
(06:21:48 PM) darkfader: Nostalgeek: yay!
(06:21:56 PM) darkfader: i didn't know if that was done already
(06:21:59 PM) ricduarte: something I find missing is memory overcommit
(06:22:37 PM) jfontan left the room (quit: Quit: Leaving.).
(06:22:49 PM) rolnas: jmelis: I also want to have sunstone extensibility, like custom tabs with some custom functionality
(06:22:51 PM) jmelis: darkfader: but sunstone 4.0 is full of wizards
(06:22:54 PM) Nostalgeek: ricduarte, did you try setting HYPERVISOR_MEM to a negative value? maybe not ideal but works great for me
(06:23:10 PM) jmelis: rolnas: all the functionality is written as plugins
(06:23:31 PM) rolnas: where is a description how to write my own plugin
(06:23:32 PM) rolnas: ?
(06:23:36 PM) jmelis: https://github.com/OpenNebula/one/tree/master/src/sunstone/public/js/plugins
(06:23:39 PM) Nostalgeek: ricduarte, http://firstname.lastname@example.org/msg10374.html
(06:23:45 PM) ricduarte: Nostalgeek, never tried, but can you control the amount of overcommit with it?
(06:23:55 PM) lars_eric left the room (quit: Quit: Leaving).
(06:23:56 PM) jmelis: not there yet, but it will be there at some point
(06:24:33 PM) darkfader: jmelis: <3
(06:24:39 PM) darkfader: i know what i'll do today
(06:24:41 PM) ricduarte: @jmelis, having vmware support without supporting memory overcommit does not look right
(06:24:58 PM) jmelis: so, back to the self-service discussion ( Nostalgeek and darkfader ) do we agree that if we add an interface to sunstone
(06:25:30 PM) jmelis: where a user can instantiate a template after modifying a few attributes
(06:25:42 PM) jmelis: like small instance template, without a disk
(06:25:45 PM) jmelis: so during instantiation
(06:25:48 PM) jmelis: you select the OS
(06:25:58 PM) Nostalgeek: ricduarte, check out that mailing-list post, its works great for me, although doesn't allow per-vm overcomming it works as a whole for all your hosts
(06:26:05 PM) jmelis: do we agree that if we bring that functionaluty to sunstone, we don't need the self-service?
(06:26:28 PM) darkfader: probably yes
(06:26:42 PM) ricduarte: @jmelis, that would fit my use cases as well
(06:26:46 PM) Nostalgeek: jmelis, I can't tell for self-service i don't use it
(06:27:01 PM) jmelis: sorry I meant rolnas, not Nostalgeek :)
(06:27:13 PM) jmelis: rolnas?
(06:27:23 PM) ricduarte: @jmelis, if we can limit a view to those wizards
(06:27:41 PM) jmelis: ricduarte: agreed
(06:27:51 PM) rolnas: I found that some problems with sunstone/opennebula could be solved with *.erb code in occi/ec2
(06:28:33 PM) rolnas: but in general I see no point to have two different GUI with same purpose
(06:28:36 PM) rolnas: almost
(06:29:13 PM) jmelis: rolnas: I understand what you're saying, but could you be a bit more specific?
(06:29:44 PM) Nostalgeek: I salute the new actions which are much more easy for a user to understand in the new Sunstone
(06:29:47 PM) rolnas: it was in past, with 4.0 I didn't yet have time to investigate more
(06:30:20 PM) jmelis: ok, if you find anything please let us know, because that's a very important issue IMO
(06:30:22 PM) ricduarte: @jmelis, something I also find missing is Cloudauth @ API bindings
(06:30:24 PM) rolnas: our users are very different, some of them are sysadmins, others more like end users
(06:31:02 PM) rolnas: so we should make solutions working for such big range of users
(06:31:29 PM) ricduarte: I can't use serveradmin without "stealing" some code from opennebula
(06:31:30 PM) jmelis: ricduarte: what do you mean by Cloudauth?
(06:31:34 PM) jmelis: you mean the OCAR ruby api?
(06:31:38 PM) jmelis: *OCA
(06:31:47 PM) ricduarte: yep, ruby
(06:32:02 PM) jmelis: ricduarte: ok, noted :)
(06:32:52 PM) rolnas: what about external interface for cloudauth ?
(06:33:51 PM) jmelis: rolnas: it's what ricduarte said, right?
(06:34:57 PM) rolnas: probably, for e.g. I would like to make ec2 work with some data in opennebula user attributes (like EC2_KEYPAIR...)
(06:35:40 PM) jmelis: that makes a lot of sense
(06:36:30 PM) jmelis: so, let's wrap up
(06:36:42 PM) jmelis: does anyone have any other topic suggestion, etc?
(06:37:23 PM) jmelis: great :)
(06:37:26 PM) jmelis: so
(06:37:44 PM) jmelis: thanks *a lot* for your comments and feedback
(06:37:55 PM) ricduarte: i would just like to add that EC2 compatibility should see more effort
(06:38:27 PM) ricduarte: there are lot of cool solutions that will not work with opennebula
(06:38:34 PM) ricduarte: and do work with competing cloud stacks
(06:38:37 PM) rolnas: lets make opennebula to take the best things from openstack :-)
(06:39:03 PM) jmelis: :)
(06:39:06 PM) ricduarte: as this is my first attendance, I would like to thanks to the opennebula dev
(06:39:29 PM) rolnas: thanks from my also
(06:39:35 PM) Nostalgeek: Yes, thanks everyone
(06:39:35 PM) ricduarte: I tried all the major stacks and no other was as easy, fas, relaaable and customizable as opennebula
(06:40:01 PM) Nostalgeek: looking forward at meeting you guys in Berlin in September! :)
(06:40:03 PM) rolnas: ricduarte: I agree
(06:40:10 PM) jmelis: brilliant!!
(06:40:17 PM) jmelis: that was going to be my closing line
(06:40:37 PM) jmelis: for anyone out there who hasn't heard... http://opennebulaconf.com :)
(06:41:07 PM) jmelis: by the way
(06:41:20 PM) jmelis: we will send (probably tomorrow) an email recapping this session
(06:41:58 PM) jmelis: so thank you guys a lot
(06:43:00 PM) ricduarte: thank you
(06:43:56 PM) jmelis has changed the topic to: OpenNebula 4.0 is out! http://opennebula.org/software:software || http://opennebulaconf.com/ 24-26 Sept, Berlin
(06:44:40 PM) Nostalgeek: oh, and Google just opened up registration for their Google Compute Engine
(06:44:47 PM) Nostalgeek: something to keep an eye on and add support in OpenNebula
(06:45:20 PM) Nostalgeek: I might be available to help if anyone interested in looking at getting GCE support
(06:45:55 PM) jmelis: btw joelio: http://dev.opennebula.org/issues/2009 this will make it to 4.0.1 and would have fixed your problem earlier this morning
(06:46:00 PM) jmelis: Nostalgeek: good to know
(06:46:22 PM) jmelis: we registered for GCE accounts when they first announce it and it's something we're definitely interested in
(06:47:49 PM) tim|mint left the room (quit: Quit: Leaving).
(06:48:28 PM) Nostalgeek: hehe ok, i tried getting in the beta back last year but they ignored :)
(06:48:41 PM) Nostalgeek: I'll be giving it a try soon