07:19:20 <tosky> #startmeeting sahara
07:19:20 <opendevmeet> Meeting started Tue Oct 19 07:19:20 2021 UTC and is due to finish in 60 minutes.  The chair is tosky. Information about MeetBot at http://wiki.debian.org/MeetBot.
07:19:20 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
07:19:20 <opendevmeet> The meeting name has been set to 'sahara'
07:19:41 <tosky> #chair qiujunting ruifaling
07:19:41 <opendevmeet> Current chairs: qiujunting ruifaling tosky
07:20:25 <tosky> welcome to the meeting! qiujunting, can you please ask again your last question ?
07:24:39 <qiujunting> Can you introduce the organization process of the regular project meeting?
07:25:39 <qiujunting> Consider the regular meeting of the Sahara project
07:27:42 <qiujunting> tosky:
07:28:27 <tosky> normal IRC meetings are more or less the same, and described here: https://wiki.openstack.org/wiki/Meetings/ChairaMeeting
07:28:41 <tosky> you go through the points on your agenda one by one
07:29:40 <qiujunting> ok, thanks
07:30:05 <qiujunting> I have a point that saharaahara support the creating the instance by specifying system volumes.
07:30:25 <qiujunting> I have a point that sahara support the creating the instance by specifying system volumes.
07:31:13 <qiujunting> The VM instance boot by a volume.
07:31:58 <ruifaling> An existing volume ?
07:32:06 <qiujunting> yes
07:33:13 <ruifaling> so you mean  there are some volumes has been created,then we lanch a cluster by those volumes ?
07:33:48 <qiujunting> First create the boot volume of the VM instance, and specify the volume boot VM instance.
07:33:59 <qiujunting> yes
07:34:39 <tosky> but sahara supports boot from volume, as far as I remember
07:35:28 <qiujunting> This can reduce the time it takes for Sahara to deploy the cluster.
07:35:29 <ruifaling> yes，
07:35:32 <tosky> it was added here: https://storyboard.openstack.org/#!/story/2001820
07:35:57 <tosky> oh, do you mean having sahara deploy on VMs which are already running?
07:39:04 <qiujunting> No, I mean it is faster for nova to start the virtual machine with volume, and Sahara can assign this volume to nova.
07:39:48 <tosky> but sahara doesn't start the volume, sahara asks nova to start virtual machines
07:40:12 <tosky> and sahara supports boot from volume, which was described by https://specs.openstack.org/openstack/sahara-specs/specs/backlog/boot-from-volume.html
07:40:31 <tosky> that document is in the "backlog" section but it was implemented by https://storyboard.openstack.org/#!/story/2001820
07:40:40 <tosky> how does your proposal differ from that?
07:44:23 <qiujunting> I mean like this：https://specs.openstack.org/openstack/sahara-specs/specs/backlog/boot-from-volume.html
07:45:05 <qiujunting> sahara has been suuported this ?
07:46:42 <tosky> yes, please see the commits referenced by https://storyboard.openstack.org/#!/story/2001820
07:55:35 <ruifaling> so the other topic to discussion in qiu's email is :Sahara deploys a dedicated cluster through cloud host VM tools (qemu-guest-agent)
07:57:19 <ruifaling> maybe it means using QGA to deploy clusters
07:57:33 <opendevreview> Ian Wienand proposed openstack/openstack-ansible-os_sahara stable/stein: Remove debian-stable jobs  https://review.opendev.org/c/openstack/openstack-ansible-os_sahara/+/814349
08:00:30 <tosky> can you please explain this point a bit more?
08:02:17 <ruifaling> In the current version, the script is executed by Sahara through SSH to the VM
08:03:17 <ruifaling> It rely on the external networks
08:05:49 <tosky> or even internal network, but yes, that's correct, the controller should be able to reach the VMs through some network
08:06:33 <ruifaling> Use qga to execute the script  could weakening the dependence on  network
08:06:55 <tosky> but this means a tight dependency between sahara and the instances
08:07:37 <ruifaling> yes
08:07:48 <tosky> sahara doesn't know about the way instances are managed, it would go against being an abstraction layer
08:07:57 <tosky> also, nova can support multiple ways of deployments
08:08:08 <tosky> and that wouldn't work with baremetal instances of course
08:11:27 <ruifaling> oh yes,BM,may be distinguish between virtual machines and bare metal machines
08:12:29 <tosky> but again, you would also need to access the hypervisor, and that solution would be targeted to a specific hypervisor
08:15:14 <tosky> moreover, the virtual machines requires to have a network anyway, because the hadoop services on the cluster need to talk to each other
08:16:56 <qiujunting> yes, before we need the falt network, by
08:17:44 <qiujunting> by qga, we need not network must be flat network.
08:18:18 <ruifaling> the controller reach to the vms not very necessary
08:20:43 <tosky> but it's the only way to not break the abstraction
08:21:16 <tosky> you don't need the network to be a flat one, because you can configure the access though a specific node using the proxy settings
08:21:30 <tosky> https://docs.openstack.org/sahara/latest/admin/advanced-configuration-guide.html#custom-network-topologies
08:24:58 <tosky> you can set up a "jumphost" which can be reached by the controllers, and that can reach the instances
08:38:08 <ruifaling> ok,I understand ,thank you,tosky
08:54:43 <tosky> qiujunting: is there any other topic before closing?
08:55:07 <qiujunting> no, thank you tosky
08:55:10 <tosky> for example, do you have a plan about which backends should be kept and updated and which ones should be dropped?
08:55:37 <tosky> sorry, plugins - I suspect some of them may not work anymore, especially ambari and cdh, given the recent changes in Cloudera
08:55:47 <tosky> and the vanilla one is very outdated
08:55:53 <tosky> do you plan any work on updating them?
08:57:40 <qiujunting> If possible, consider updating them in the yoga version.
08:58:21 <qiujunting> You can introduce the plan in detail.
08:59:16 <qiujunting> We can communicate by email
09:05:23 <tosky> I don't have a plan because my time on sahara is limited, I can help with revisions
09:05:45 <tosky> but right now the supported plugins are not really useful for the users
09:05:57 <tosky> so if you want to use it, you may be interested in updating them as well
09:06:05 <tosky> anything else? Can I close the meeting?
09:06:36 <tosky> (please note that I set you both as chair for this specific meeting with the command #chair , so anyone of us can close the meeting with #endmeeting )
09:08:50 <tosky> qiujunting, ruifaling ^
09:09:19 <ruifaling> ok，thank you very much
09:09:27 <ruifaling> bye
09:09:33 <tosky> thank you all
09:09:38 <tosky> #endmeeting