15:00:17 <anteaya> #startmeeting third-party
15:00:18 <openstack> Meeting started Mon Feb 16 15:00:17 2015 UTC and is due to finish in 60 minutes.  The chair is anteaya. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:19 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:21 <openstack> The meeting name has been set to 'third_party'
15:00:32 <anteaya> say hello if you are here for the third party meeting
15:08:18 <wznoinsk_> i
15:08:22 <wznoinsk_> hi
15:09:54 <anteaya> hello wznoinsk
15:09:58 <anteaya> nice to see you
15:10:10 <anteaya> how are things going for you this week?
15:13:48 <wznoinsk> hi anteaya, good to see you too, I'm back online after a few days off sick, I'm getting to my NFV CI shortly
15:14:09 <anteaya> good to hear you are feeling better
15:14:17 <anteaya> sorry to hear you were feeling ill
15:14:21 <wznoinsk> I think there was a bit of confusion about our NFV CI as a) it was evolving, b) new people joined
15:14:26 <anteaya> yes
15:14:27 <wznoinsk> I do apologize for that
15:14:31 <anteaya> sounds like on both sides
15:14:39 <anteaya> no problem, you can only convey what you know
15:14:53 <anteaya> and I think part of this exercise helped folks to increase what it is you know
15:15:02 <anteaya> so that is a net benefit for all
15:15:58 <anteaya> anything you wanted to talk about today?
15:16:16 <anteaya> you did see that gerrit will be changing ip addresses on March 21, yeah?
15:16:28 <wznoinsk> that's true, yes I saw that
15:16:33 <anteaya> great
15:16:58 <anteaya> so you are aiming to test nova repos correct?
15:17:08 <anteaya> are you expecting to test any other project repos?
15:17:46 <wznoinsk> we're testing neutron project with our networking CI already, that's to test our dpdk-ovs as part of NFV CI
15:18:01 <wznoinsk> I hit issues with docker last week but it should be back to commenting today
15:18:05 <anteaya> okay so both neutron and nova repos
15:18:11 <anteaya> you are testing docker?
15:18:37 <wznoinsk> the other part (for nova) is going to test only on nova and apart from standard tempest nova tests we're going to have custom on the new features (cpu pinning, huge pages etc.)
15:19:25 <wznoinsk> I'm using docker as the management tool to deploy containers for jenkins builds/slaves, I'm not using docker inside the openstack
15:19:37 <anteaya> oh
15:19:47 <anteaya> I still have no idea what huge pages are
15:20:49 <wznoinsk> the problem with using VMs for CI systems is that the nwe and hot hardware/kernel features are not emulated inside the VMs so no way to test them
15:21:17 <wznoinsk> kernel features relying on hardware*
15:21:31 <anteaya> okay
15:22:03 <wznoinsk> with containers you can have a full access to the host's hardware and it's capabilities
15:23:30 <anteaya> oh
15:23:39 <anteaya> I can understand that then
15:23:44 <anteaya> what are huge pages
15:24:05 <wznoinsk> with regards to hugepages it's giving the same amount memory to an application/vm/container as it requested by a memory page (a unit of memory) is larger, hence memory is less fragmented and less memory page lookups is needed during processing
15:24:24 <wznoinsk> s/by/but
15:25:03 <anteaya> interesting
15:25:16 <anteaya> again that would need access to hardware yes?
15:25:21 <anteaya> I don't see vms capable of that
15:26:50 <wznoinsk> if an application need 500MB of memory, instead of 125000 memory pages of size 4k (default for most OSes) an os/programmer/sysadmin can assign huge pages of 2MB so it will allocated only 250 (2MB x 250 = 500MB)
15:27:18 <wznoinsk> it's access to memory, kernels fully support hugepages already
15:27:24 <wznoinsk> grep -i huge /proc/meminfo
15:27:36 <anteaya> cool
15:29:37 <wznoinsk> it's quite an interesteing piece from the series how to max out the performance of your apps (interest of NFV)
15:30:07 <kashyap> anteaya, Randomly chiming in here - you can look up info on HugePages here: https://wiki.debian.org/Hugepages
15:30:29 <anteaya> wznoinsk: yes it does sound interesting
15:30:32 <anteaya> kashyap: thank you
15:30:58 <kashyap> A simple test with libvirt (which is used by Nova under the hood) and huge pages - https://kashyapc.fedorapeople.org/virt/test-hugepages-with-libvirt.txt
15:32:26 <anteaya> I guess the name large pages is also used
15:34:14 <kashyap> Yes
15:34:33 <anteaya> cool
15:34:39 <anteaya> thanks for letting me know
15:34:42 <anteaya> I appreciate it
15:34:44 <anteaya> :)
15:36:52 <patrickeast_> hello, sry i’m a bit late… is the third-party meeting still going on?
15:37:06 <anteaya> hello patrickeast_
15:37:08 <anteaya> yes
15:37:16 <patrickeast_> anteaya: hey
15:37:28 <anteaya> we were talking about gerrit's ip change coming up on March 21, you saw that yeah?
15:37:38 <anteaya> and also we were talking about huge pages
15:37:55 <anteaya> which I didn't know anything about and which wznoinsk and kashyap were helping me to understand
15:38:04 <patrickeast_> yep, hopefully the ip change shouldn’t be an issue for me… our firewall is pretty friendly
15:38:14 <anteaya> it improves performance through configuring memory allocation
15:38:20 <anteaya> did I get that right?
15:38:21 <kashyap> I accidentally noticed the message and randomly chimed in :-)
15:38:30 <anteaya> patrickeast_: yay for friendly firewalls
15:38:37 <anteaya> kashyap: I'm glad you did, thank you
15:38:45 <anteaya> patrickeast_: anything new for you this week?
15:38:54 <patrickeast_> nah, not really
15:39:07 <patrickeast_> i got burned on something new but fixed it
15:39:33 <patrickeast_> turns out nodepool image caching doesn’t work if your openstack provider network on the template instance has a 10.0.0.0/24 subnet
15:39:42 <wznoinsk> anteaya: yes, spot on
15:39:44 <anteaya> good to know
15:39:50 <anteaya> wznoinsk: yay!
15:39:51 <patrickeast_> had to shut everything down and move it to a different one
15:39:56 <anteaya> thanks to good teaching
15:40:04 <anteaya> patrickeast_: what did you pick?
15:40:19 <patrickeast_> 10.0.10.0/24
15:40:30 <patrickeast_> anything should work except for 10.0.0.0
15:41:24 <patrickeast_> more info here https://bugs.launchpad.net/devstack/+bug/1418250
15:41:26 <openstack> Launchpad bug 1418250 in devstack "image_list fails when host_ip falls in FIXED_RANGE" [Undecided,In progress] - Assigned to Patrick East (patrick-east)
15:42:26 <anteaya> #info for nodepool image caching using anything but 10.0.0.0 subnet on your openstack provider network template instance
15:42:33 <anteaya> patrickeast_: did I get that right?
15:43:01 <patrickeast_> yea
15:43:31 <anteaya> great
15:43:36 <anteaya> thanks for sharing that
15:43:49 <patrickeast_> maybe provider network for nodepool template instance would make it more clear, but i think folks setting this up will know what that means
15:44:29 <anteaya> #info better wording: provider network for nodepool template instance
15:44:35 <anteaya> patrickeast_: better?
15:44:39 <patrickeast_> perfect
15:44:41 <anteaya> thanks
15:44:43 <anteaya> :)
15:44:57 <anteaya> anything anyone else wants to talk about today?
15:45:49 <patrickeast_> nothin more from me
15:46:11 <anteaya> cool
15:48:42 <anteaya> the only other thing I can think of is making sure you are aware of asselin's spec patch: https://review.openstack.org/#/c/139745/
15:48:58 <anteaya> this is the follow up work that the puppet module split out was a foundation for
15:49:06 <anteaya> don't know if you know about it
15:49:14 <anteaya> but it might be an idea to take a look
15:51:08 <patrickeast_> anteaya: thanks for the link, i know this has been in the works but i keep forgetting to actually go review the spec
15:51:21 <anteaya> go review the spec
15:51:36 <anteaya> and actually asselin should add it to an upcoming infra meeting agenda
15:51:49 <anteaya> so it gets on the priority list if there is a spot for it
15:57:56 <anteaya> and we are coming up to the end of our time
15:58:06 <anteaya> so I'll close it off here
15:58:15 <anteaya> thanks everyone, nice to talk to you today
15:58:19 <anteaya> see you next week
15:58:22 <anteaya> #endmeeting