| *** jtriley has quit IRC | 00:07 | |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Allow specifying erasure as the pool type https://review.openstack.org/236754 | 00:08 |
|---|---|---|
| openstackgerrit | Sam Yaple proposed openstack/kolla: Allow ceph pool names to be configurable https://review.openstack.org/236755 | 00:08 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Automate setting up a cache tier https://review.openstack.org/236756 | 00:08 |
| *** dolpher has joined #kolla | 00:14 | |
| *** dolpher has quit IRC | 00:16 | |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add additional ntp information around ceph https://review.openstack.org/236758 | 00:30 |
| *** jasonsb has joined #kolla | 00:44 | |
| *** dolpher has joined #kolla | 00:44 | |
| *** tpot has joined #kolla | 00:45 | |
| *** dolpher has quit IRC | 00:45 | |
| *** Kennan has quit IRC | 00:46 | |
| *** Kennan has joined #kolla | 00:46 | |
| *** vinkman has joined #kolla | 00:46 | |
| *** dims_ has joined #kolla | 00:53 | |
| *** dimsum__ has quit IRC | 00:56 | |
| *** dolpher has joined #kolla | 00:59 | |
| *** dims_ has quit IRC | 01:03 | |
| *** jtriley has joined #kolla | 01:03 | |
| *** dimsum__ has joined #kolla | 01:05 | |
| *** dimsum__ has quit IRC | 01:08 | |
| *** dimsum__ has joined #kolla | 01:08 | |
| *** jtriley has quit IRC | 01:09 | |
| asalkeld | guys are the summit sessions figured out yet? | 01:28 |
| asalkeld | hard to plan atm, as i don't know what's in each session | 01:28 |
| *** achanda has quit IRC | 01:35 | |
| *** zehicle has quit IRC | 01:55 | |
| *** zehicle has joined #kolla | 01:59 | |
| *** jtriley has joined #kolla | 02:05 | |
| *** jtriley has quit IRC | 02:10 | |
| *** achanda has joined #kolla | 02:19 | |
| *** dimsum__ has quit IRC | 02:24 | |
| *** dimsum__ has joined #kolla | 02:24 | |
| *** CBR09 has joined #kolla | 02:32 | |
| *** klint has joined #kolla | 02:40 | |
| *** dimsum__ has quit IRC | 02:51 | |
| *** dimsum__ has joined #kolla | 02:51 | |
| *** jtriley has joined #kolla | 03:06 | |
| *** jtriley has quit IRC | 03:11 | |
| *** tpot has quit IRC | 03:20 | |
| *** dimsum__ has quit IRC | 03:30 | |
| *** dimsum__ has joined #kolla | 03:31 | |
| *** dimsum__ has quit IRC | 03:31 | |
| *** dimsum__ has joined #kolla | 03:31 | |
| *** dimsum__ has quit IRC | 03:37 | |
| *** dimsum__ has joined #kolla | 03:38 | |
| *** alisonh has quit IRC | 03:50 | |
| *** alisonh has joined #kolla | 03:55 | |
| *** dolpher has quit IRC | 03:56 | |
| *** dimsum__ has quit IRC | 03:57 | |
| *** daneyon has joined #kolla | 04:05 | |
| *** daneyon_ has joined #kolla | 04:06 | |
| *** jtriley has joined #kolla | 04:07 | |
| *** zehicle has quit IRC | 04:07 | |
| *** daneyon has quit IRC | 04:10 | |
| *** jtriley has quit IRC | 04:12 | |
| *** daneyon_ has quit IRC | 04:14 | |
| *** unicell1 has joined #kolla | 04:25 | |
| *** unicell has quit IRC | 04:26 | |
| *** tpot has joined #kolla | 04:39 | |
| *** bmace has quit IRC | 04:40 | |
| *** bmace has joined #kolla | 04:40 | |
| *** jtriley has joined #kolla | 05:08 | |
| *** asalkeld has quit IRC | 05:09 | |
| *** asalkeld has joined #kolla | 05:10 | |
| *** tfukushima has joined #kolla | 05:11 | |
| *** jtriley has quit IRC | 05:14 | |
| *** vinkman has quit IRC | 05:28 | |
| *** tpot_ has joined #kolla | 05:33 | |
| *** tpot has quit IRC | 05:34 | |
| *** dolpher has joined #kolla | 05:42 | |
| *** tummy has joined #kolla | 05:42 | |
| *** shardy_a1k has joined #kolla | 05:52 | |
| *** shardy_afk has quit IRC | 05:54 | |
| *** dimsum__ has joined #kolla | 05:54 | |
| *** shardy_a1k has quit IRC | 05:57 | |
| *** shardy_afk has joined #kolla | 05:58 | |
| *** dimsum__ has quit IRC | 05:59 | |
| nihilifer | good morning | 06:00 |
| *** tpot_ has quit IRC | 06:09 | |
| *** jtriley has joined #kolla | 06:10 | |
| *** tpot has joined #kolla | 06:12 | |
| *** jtriley has quit IRC | 06:15 | |
| *** tpot has quit IRC | 06:17 | |
| *** sdake has joined #kolla | 06:32 | |
| sdake | evening dudes | 06:41 |
| sdake | sorry was in a different state this weekend | 06:41 |
| sdake | so no tag has happened | 06:42 |
| sdake | the reson i haven ttagged is because when i upgraded to my 1.2 tb pcie-ssd i temporarily lost my signing keys | 06:42 |
| sdake | i could make new ones | 06:42 |
| sdake | perhaps i'll just do that | 06:42 |
| sdake | SamYaple make sense? | 06:42 |
| *** CBR09 has quit IRC | 06:42 | |
| *** achanda has quit IRC | 06:44 | |
| *** unicell has joined #kolla | 06:55 | |
| *** unicell1 has quit IRC | 06:56 | |
| *** dimsum__ has joined #kolla | 06:56 | |
| *** dimsum__ has quit IRC | 07:00 | |
| *** jtriley has joined #kolla | 07:11 | |
| *** tummy has quit IRC | 07:14 | |
| *** jtriley has quit IRC | 07:16 | |
| *** akwasnie has joined #kolla | 07:25 | |
| *** jmccarthy has quit IRC | 07:31 | |
| *** jmccarthy has joined #kolla | 07:32 | |
| *** shardy_afk is now known as shardy | 07:36 | |
| *** inc0 has joined #kolla | 07:44 | |
| inc0 | good morning | 07:45 |
| *** gfidente has joined #kolla | 07:51 | |
| *** athomas has joined #kolla | 07:54 | |
| *** sdake has quit IRC | 07:56 | |
| *** jtriley has joined #kolla | 08:12 | |
| *** mbound has joined #kolla | 08:13 | |
| *** jtriley has quit IRC | 08:17 | |
| *** shardy_ has joined #kolla | 08:25 | |
| *** shardy has quit IRC | 08:26 | |
| kjelly | Hi, This is my first time to go openstack summit. I have one question. What is work session in kolla ? | 08:29 |
| *** shardy_ has quit IRC | 08:30 | |
| inc0 | kjelly, we sit in same room and discuss problem at hand | 08:31 |
| *** shardy has joined #kolla | 08:31 | |
| inc0 | there are 2 types of sessions: working and fishbowl | 08:31 |
| inc0 | working is sitting and talking, fishblowl is one person talking and rest listens | 08:32 |
| kjelly | inc0: Ok. Thanks for your answering. | 08:34 |
| inc0 | guys, shouldn't we add our stuff here as well? https://wiki.openstack.org/wiki/Design_Summit/Mitaka/Etherpads | 08:34 |
| SamYaple | asalkeld: all of our sessions are scheduled, yes | 08:36 |
| SamYaple | morning inc0 | 08:36 |
| inc0 | hey SamYaple, asalkeld are you going to commit to kolla?;) | 08:36 |
| inc0 | SamYaple, our sessions are scheduled? | 08:37 |
| inc0 | where are they? | 08:37 |
| SamYaple | yea that happened a while ago i believe | 08:37 |
| inc0 | I remember writing them down, but never seen actual schedule | 08:38 |
| *** pbourke has quit IRC | 08:40 | |
| *** pbourke has joined #kolla | 08:40 | |
| SamYaple | i might be wrong. im sure sdake has that info | 08:41 |
| SamYaple | inc0: got some patches in the queue you might like | 08:41 |
| SamYaple | rabbitmq-clusterer and ceph stuff | 08:41 |
| inc0 | I've seen, ceph stuff? | 08:41 |
| inc0 | looks good! | 08:41 |
| SamYaple | yea the cache stuff builds but glance says writing through the cache is readonly | 08:42 |
| SamYaple | still trying to figure that out | 08:42 |
| SamYaple | ive only done the cache with nova before | 08:42 |
| SamYaple | maybe glance/cinder cant use the cache? seems wierd though | 08:42 |
| inc0 | it should be transparent for them | 08:43 |
| inc0 | it's on block device layer right? | 08:44 |
| SamYaple | i agree it should be transparent, and yet once i set the overlay it stops being able to write to it | 08:45 |
| *** mbound has quit IRC | 08:47 | |
| inc0 | hmm...so our reference conf for ceph is cache and journal on this cached fs? | 08:47 |
| SamYaple | or reference right now is journals collocated on OSDs | 08:47 |
| inc0 | ok | 08:49 |
| inc0 | hmm | 08:49 |
| inc0 | but it only makes sense if you have cache | 08:49 |
| inc0 | on ssd | 08:49 |
| SamYaple | i wouldn't say that. there was never a requirement for journals on ssds | 08:49 |
| inc0 | ow...by the way...if we have writeback cache on ssd...why do we need journal for? | 08:49 |
| SamYaple | its just better performance | 08:49 |
| inc0 | it is a requirement if you want to have any performance on ceph | 08:50 |
| SamYaple | journal is always needed. whether colocated in a file or partition, or on a seperate disk. journal is required | 08:50 |
| inc0 | I mean technically you can have journal on spinning disks...you just don't want it | 08:50 |
| SamYaple | inc0: there are several people i know that have all ssd clusters now | 08:50 |
| SamYaple | wrong | 08:50 |
| SamYaple | thats not how journals work | 08:50 |
| inc0 | yeah, all sdd is cool | 08:50 |
| SamYaple | you must have a journal. period | 08:51 |
| inc0 | all spinning...having at least ssd journal is preferable | 08:51 |
| SamYaple | sure, but far from required | 08:51 |
| SamYaple | you wont get worse performance than say single node cinder | 08:51 |
| inc0 | yeah, I know | 08:51 |
| SamYaple | anyway, long story short right now with this reference arch you can hotswap journals between hosts and kolla will pick up on that | 08:52 |
| SamYaple | it wont update teh crush map _but_ it will start the osd on the new host | 08:52 |
| inc0 | yeah, I like this approach anyway | 08:52 |
| SamYaple | with external journals that isnt possible | 08:52 |
| *** dimsum__ has joined #kolla | 08:59 | |
| SamYaple | external journals are going to be wierd | 09:03 |
| *** dimsum__ has quit IRC | 09:04 | |
| *** mbound has joined #kolla | 09:05 | |
| *** openstack has joined #kolla | 09:19 | |
| *** SamYaple has quit IRC | 09:20 | |
| *** SamYaple has joined #kolla | 09:20 | |
| *** dolpher has quit IRC | 09:21 | |
| *** SamYaple has quit IRC | 09:26 | |
| *** SamYaple has joined #kolla | 09:26 | |
| *** SamYaple has quit IRC | 09:26 | |
| *** SamYaple has joined #kolla | 09:27 | |
| *** akwasnie has quit IRC | 09:49 | |
| *** mandre has quit IRC | 09:50 | |
| *** mandre has joined #kolla | 09:51 | |
| *** dimsum__ has joined #kolla | 10:07 | |
| *** mandre has quit IRC | 10:11 | |
| *** athomas has quit IRC | 10:12 | |
| *** mandre has joined #kolla | 10:12 | |
| *** jtriley has joined #kolla | 10:13 | |
| *** athomas has joined #kolla | 10:17 | |
| *** jtriley has quit IRC | 10:18 | |
| *** akwasnie has joined #kolla | 10:32 | |
| asalkeld | SamYaple: i'd like to see the schedule as there are lots of conflicts with heat | 10:39 |
| asalkeld | (to see what each session is about) | 10:39 |
| inc0 | asalkeld, I don't think we have schedule yet | 10:39 |
| inc0 | I'll ask sdake when he'll come back | 10:40 |
| asalkeld | cool, thanks | 10:40 |
| asalkeld | inc0: yeah i want to get into kolla, just playing with it atm | 10:40 |
| inc0 | cool:) it would be great to have you here | 10:40 |
| asalkeld | building images is a real pita | 10:41 |
| inc0 | yeah? why is that? | 10:41 |
| *** dolpher has joined #kolla | 10:41 | |
| asalkeld | mine fails most of the time | 10:41 |
| asalkeld | network mirror failures | 10:41 |
| inc0 | well we are bound to repos | 10:42 |
| asalkeld | (maybe being in AU - tho' not sure) | 10:42 |
| inc0 | I personally use ubuntu-source, this seems most stable | 10:42 |
| inc0 | try centos-source, delorean causes most of build problems | 10:42 |
| inc0 | also make sure that you use --no-cache every now and then, especially with delorean | 10:43 |
| asalkeld | k | 10:43 |
| inc0 | because docker caches repo urls and these tend to disapear | 10:43 |
| *** tfukushima has quit IRC | 10:43 | |
| inc0 | by delorean I mean binary building | 10:44 |
| asalkeld | so base=ubuntu, install_type=source ? | 10:45 |
| inc0 | yup | 10:45 |
| inc0 | or tools/build.py --type source --base ubuntu | 10:45 |
| inc0 | I see that our gates currently fails to build designate, probably dependency fuckup, but you don't need designate to deploy aio | 10:46 |
| asalkeld | trying again now... | 10:46 |
| inc0 | let me know if you run to any problems | 10:46 |
| asalkeld | thanks | 10:47 |
| inc0 | also remember to change globals.yml to deploy ubuntu source | 10:47 |
| *** dolpher has quit IRC | 10:48 | |
| inc0 | RequirementParseError: Missing distribution spec /designate as I said...repo roblems :/ | 10:48 |
| SamYaple | asalkeld: if you want to speed it up you can limit what you build. this will build all images that work with ansible `~/kolla/tools/build.py --base ubuntu --type source ceph cinder data glance haproxy heat horizon keepalived keystone kolla-ansible rsyslog mariadb memcached neutron nova openvswitch rabbitmq murano ironic` | 10:50 |
| SamYaple | asalkeld: pbourke has a patch upstream to make that cleaner with build profiles, but it does cut the build time in half | 10:50 |
| *** dimsum__ is now known as dims | 10:52 | |
| inc0 | SamYaple, I'm waiting for machines to test out your ceph stuff | 10:55 |
| inc0 | this really looks cool, but I'd like to see it running before I approve it | 10:55 |
| inc0 | also, what's up with udev? did we make any progress on that front? | 10:56 |
| asalkeld | thanks SamYaple | 10:57 |
| inc0 | also, SamYaple is there reason why we use some strange github accounts here? https://review.openstack.org/#/c/236669/2/etc/kolla/kolla-build.conf | 11:02 |
| inc0 | also novnc seems....not very openstack-ish | 11:02 |
| SamYaple | inc0: yea but i can't backport the udev stuff so im still thinking it out | 11:02 |
| SamYaple | inc0: that is infact what openstack uses :) | 11:02 |
| SamYaple | like offically | 11:03 |
| inc0 | meh, fork would be good | 11:03 |
| inc0 | ok..lol | 11:03 |
| SamYaple | its not an openstack project though | 11:03 |
| inc0 | I didn't know that | 11:03 |
| SamYaple | if you check all the packaging youll see thats where the source is from | 11:03 |
| SamYaple | for novnc and spice | 11:03 |
| inc0 | so technically one random guy called kanaka can break openstack? with devstack and such? | 11:03 |
| inc0 | yay \o/ | 11:04 |
| SamYaple | well no. i doubt devstack slones from the novnc repo directly | 11:04 |
| SamYaple | thats what the packages are built on | 11:04 |
| SamYaple | nvm | 11:05 |
| SamYaple | they do | 11:05 |
| SamYaple | https://github.com/openstack-dev/devstack/blob/b2b82a0f34d84ecf5329703bf6e63c89272ef2da/stackrc#L498 | 11:05 |
| inc0 | yeah, I was looking for exact line, I knwo they do | 11:06 |
| inc0 | also look - spice has different url than we use | 11:06 |
| SamYaple | yea but the github url is the offical | 11:06 |
| inc0 | someone should probably fix devstack then | 11:07 |
| SamYaple | besides ill trust github to stay up over freedesktop.org | 11:07 |
| SamYaple | nah they are both the same its fix | 11:07 |
| SamYaple | i mean you can do what you want | 11:07 |
| inc0 | as always, but defaults matter | 11:07 |
| SamYaple | inc0: oh look at that | 11:11 |
| SamYaple | it changed | 11:11 |
| SamYaple | they are cloning the right place | 11:11 |
| SamYaple | all new commit go there | 11:11 |
| SamYaple | it changed about 6 months ago | 11:11 |
| inc0 | that's a nope https://github.com/openstack-dev/devstack/blob/master/stackrc#L502 | 11:12 |
| SamYaple | what? | 11:12 |
| inc0 | they still clone freedesktop | 11:12 |
| inc0 | on spice | 11:13 |
| SamYaple | no im saying spice is no longer commiting to github | 11:13 |
| inc0 | ah | 11:13 |
| SamYaple | its all done on freedesktop.org | 11:13 |
| SamYaple | as of about 6 months ago it seems | 11:13 |
| inc0 | so we clone from wrong place | 11:13 |
| SamYaple | well none of that code is commited so we dont clone from anywher yet | 11:13 |
| inc0 | https://www.youtube.com/watch?v=u28oYid7pOU | 11:13 |
| inc0 | sure, I mean we should | 11:14 |
| inc0 | anyway, let me -1 it:P | 11:14 |
| *** jtriley has joined #kolla | 11:14 | |
| inc0 | unless you want to do the honors? | 11:14 |
| SamYaple | im not sure im ok with using freedesktop. github is the official one from spine | 11:15 |
| SamYaple | spice* | 11:15 |
| SamYaple | freedesktop has ~8 additional commits from redhat | 11:15 |
| *** rhallisey has joined #kolla | 11:15 | |
| SamYaple | im thinking theres something else up here | 11:15 |
| inc0 | meh, is github code stable? | 11:17 |
| SamYaple | since well be calling a tag its identical code | 11:17 |
| SamYaple | still trying to figure out the disconnect here | 11:18 |
| inc0 | probably they noticed that freedesktop is more active and didn't pay much more attention to it | 11:18 |
| *** jtriley has quit IRC | 11:19 | |
| SamYaple | it looks like spice-devel is all through freedesktop | 11:21 |
| SamYaple | im guessing they just haven't pulled down the latest code | 11:21 |
| inc0 | maybe they use github just for releases? | 11:21 |
| SamYaple | thats possible. sync when they tag | 11:22 |
| SamYaple | the last commit was a tag | 11:22 |
| SamYaple | i think we should keep at at github. more stable, provides tarballs | 11:23 |
| SamYaple | someone else can always change it | 11:23 |
| SamYaple | ill put a note in there with the other repo | 11:23 |
| inc0 | yeah, just comment on there and let's move on | 11:24 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add nova-spicehtml5proxy console https://review.openstack.org/236669 | 11:26 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Ansible config for Spice console https://review.openstack.org/236670 | 11:26 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Update requirements to use tarball by default https://review.openstack.org/236914 | 11:28 |
| *** akwasnie has quit IRC | 12:06 | |
| *** inc0 has quit IRC | 12:07 | |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add nova-spicehtml5proxy console https://review.openstack.org/236669 | 12:09 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Ansible config for Spice console https://review.openstack.org/236670 | 12:09 |
| *** jtriley has joined #kolla | 12:15 | |
| openstackgerrit | Sam Yaple proposed openstack/kolla: add "registry" flag to "tools/build.py" https://review.openstack.org/236938 | 12:17 |
| *** diogogmt has quit IRC | 12:19 | |
| *** jtriley has quit IRC | 12:20 | |
| *** dims has quit IRC | 12:27 | |
| *** dims has joined #kolla | 12:28 | |
| openstackgerrit | Paul Bourke proposed openstack/kolla: Add build profiles to build.py https://review.openstack.org/234229 | 12:31 |
| *** inc0 has joined #kolla | 12:34 | |
| *** achanda has joined #kolla | 12:39 | |
| *** achanda has quit IRC | 12:41 | |
| *** diogogmt has joined #kolla | 12:45 | |
| *** klint has quit IRC | 12:56 | |
| SamYaple | wierd inc0 | 12:59 |
| SamYaple | i cant get python-rbd to create an rbd at all when it has the cache overlay set | 12:59 |
| SamYaple | i mean nova seems to work still (thats all i tested before) | 13:00 |
| SamYaple | but strange | 13:00 |
| inc0 | can you do it by hand? | 13:01 |
| SamYaple | i can do it with the rbd tools which are not usingthe python libs | 13:02 |
| inc0 | and can you do it by python-lib? like simple script to do it? | 13:02 |
| SamYaple | im working on seting that up | 13:03 |
| SamYaple | the wierd this is the librbd.so is what actually handles all of the reading of the configs and keys | 13:03 |
| SamYaple | so it should work..... | 13:03 |
| inc0 | do you have set of commands you use for that in normal librados cli? | 13:03 |
| inc0 | librbd even | 13:03 |
| SamYaple | rbd -p images create test --size 1024 | 13:03 |
| SamYaple | works fine | 13:03 |
| SamYaple | images is cached by images-cache | 13:04 |
| inc0 | yeah I find it unlikely | 13:04 |
| *** inc0 has quit IRC | 13:11 | |
| *** jtriley has joined #kolla | 13:15 | |
| *** diogogmt has quit IRC | 13:19 | |
| *** jtriley has quit IRC | 13:20 | |
| *** mandre has quit IRC | 13:25 | |
| *** mandre has joined #kolla | 13:25 | |
| *** masterbound has joined #kolla | 13:28 | |
| *** mbound has quit IRC | 13:29 | |
| *** exploreshaifali has joined #kolla | 13:47 | |
| *** jtriley has joined #kolla | 14:16 | |
| *** shardy_ has joined #kolla | 14:20 | |
| *** jtriley has quit IRC | 14:21 | |
| *** shardy has quit IRC | 14:22 | |
| *** sdake has joined #kolla | 14:24 | |
| *** shardy_ has quit IRC | 14:26 | |
| *** shardy has joined #kolla | 14:27 | |
| *** akwasnie has joined #kolla | 14:30 | |
| *** akwasnie has left #kolla | 14:30 | |
| SamYaple | still a problem inc0 | 14:30 |
| *** diogogmt has joined #kolla | 14:32 | |
| *** achanda has joined #kolla | 14:32 | |
| *** Hui has joined #kolla | 14:32 | |
| Hui | Morning, SamYaple | 14:32 |
| SamYaple | morning Hui | 14:34 |
| *** vinkman has joined #kolla | 14:36 | |
| Hui | SamYaple, does this patch work for you regarding this bug https://bugs.launchpad.net/kolla/+bug/1504606 ? | 14:37 |
| openstack | Launchpad bug 1504606 in kolla "Horizon instance console fails to connect instance with more than one controller" [Critical,Fix committed] - Assigned to Sam Yaple (s8m) | 14:38 |
| SamYaple | Hui: yes, on a fresh build | 14:38 |
| Hui | have you tested with 2 controllers? | 14:38 |
| SamYaple | yes | 14:38 |
| Hui | ok, let me do a fresh build | 14:38 |
| *** achanda has quit IRC | 14:40 | |
| *** Kennan2 has joined #kolla | 14:41 | |
| *** Kennan has quit IRC | 14:41 | |
| *** cemason has quit IRC | 14:41 | |
| *** cemason has joined #kolla | 14:42 | |
| *** Slower_ is now known as Slower | 14:47 | |
| SamYaple | sdake built the automated writeback cache stuff for ceph | 14:47 |
| sdake | samyaple fantastic! | 14:53 |
| sdake | samyaple this week we need to get together and do slide dev and runtrhough | 14:54 |
| sdake | it takes about 4-8 hours | 14:54 |
| SamYaple | yea | 14:54 |
| *** tfukushima has joined #kolla | 14:57 | |
| sdake | SamYaple is the cache work in master | 14:58 |
| SamYaple | i have an updated patchset im about to push | 14:59 |
| sdake | docs coming too? | 14:59 |
| SamYaple | probably not | 14:59 |
| SamYaple | parted /dev/nvme0n1 -s -- mklabel gpt mkpart KOLLA_CEPH_OSD_CACHE_BOOTSTRAP 1 -1 | 15:00 |
| SamYaple | thats all you _have_ to do | 15:00 |
| *** masterbound has quit IRC | 15:00 | |
| sdake | got it | 15:00 |
| SamYaple | the other options should be explanitory | 15:00 |
| *** mbound has joined #kolla | 15:00 | |
| *** dwalsh has joined #kolla | 15:00 | |
| sdake | can i use nvme0n2 | 15:04 |
| sdake | my nvme0n1 is my root filesystem and docker partotition | 15:05 |
| SamYaple | nope | 15:06 |
| SamYaple | oh wait nvme0n2 thats a seperate card, yes? | 15:07 |
| SamYaple | you can use a seperate full device | 15:07 |
| SamYaple | you cant just pick a partition, no | 15:07 |
| sdake | same card | 15:09 |
| SamYaple | nope | 15:09 |
| sdake | is that hard to implement, seems like a major use case to me | 15:10 |
| *** daneyon has joined #kolla | 15:10 | |
| sdake | i know you dont want to pimp out the implementation | 15:10 |
| sdake | but the basics wouldbe nice :) | 15:10 |
| SamYaple | oh you can do it if you do it manually | 15:10 |
| *** diogogmt has quit IRC | 15:10 | |
| SamYaple | the issue is with partitioning | 15:10 |
| SamYaple | it needs to be specific | 15:10 |
| sdake | i'm going to be recommending eveyrone run with a pcie inte l750 for their deployments on each node | 15:11 |
| *** daneyon_ has joined #kolla | 15:11 | |
| SamYaple | jokingly? | 15:11 |
| sdake | if you want speed to run with a 750 | 15:11 |
| sdake | yes i will state that | 15:11 |
| sdake | at summit | 15:11 |
| *** diogogmt has joined #kolla | 15:11 | |
| SamYaple | if i also say it do you think intel will buy me things? | 15:11 |
| sdake | highly unlikely :) | 15:12 |
| sdake | "come with me, if you want to live." | 15:12 |
| SamYaple | i wouldnt suggest you talk to much about reference architectures for ceph yu havent tested | 15:12 |
| sdake | makes sense | 15:13 |
| sdake | so what is the problem with labeling a second partition on a nvme device? | 15:13 |
| SamYaple | i reformat the whole device to bootstrap it | 15:13 |
| SamYaple | it expects specific names and partition setups | 15:13 |
| SamYaple | journal needs to be the second partition and osd needs to be the first | 15:14 |
| sdake | what is osd? | 15:14 |
| *** daneyon has quit IRC | 15:14 | |
| sdake | oh nevermind i get it | 15:14 |
| sdake | so basically you need two extra parittions? | 15:14 |
| sdake | beyond the 1 needed for docker + the baes os? | 15:15 |
| SamYaple | oh definetely there | 15:15 |
| SamYaple | ceph can use just raw directories but it does not work at _all_ with the implementation we have | 15:15 |
| SamYaple | raw directories are the edge case | 15:15 |
| sdake | it would just be nice if the partition is not labeled that it wasn't nuked | 15:17 |
| *** jtriley has joined #kolla | 15:17 | |
| SamYaple | the whole point is that is nuking the entire disk | 15:17 |
| SamYaple | its not using you partition | 15:17 |
| SamYaple | its using the DISK | 15:17 |
| SamYaple | that partition marker is just a flag that says "wipe this whole disk' | 15:18 |
| sdake | its a shame the nvme cannot be shared between base os and ceph cache | 15:22 |
| *** jtriley has quit IRC | 15:22 | |
| sdake | i guess i'll ha eto do it manually | 15:22 |
| SamYaple | i have it as a todo, but in the end youre still going to have to reformat that disk | 15:23 |
| sdake | the problem with that model is if the system has two disks, everything is used up by ceph | 15:23 |
| SamYaple | i frankly dont care about that. 2 disks is not what im targeting nor is it an accurate repreentation of ceph | 15:23 |
| sdake | i think its likely people will have a system with 1 ceph disk and 1 storage cahe, dont you? | 15:26 |
| sdake | and no extra storage for the os+docker dir? | 15:26 |
| SamYaple | 1 ceph osd per host? no | 15:27 |
| SamYaple | i know no one like that | 15:27 |
| SamYaple | 4-20 is what i see | 15:27 |
| sdake | no, one storage cache per host | 15:27 |
| SamYaple | sure | 15:27 |
| SamYaple | but they arent sharing it | 15:27 |
| SamYaple | thats a horrible idea | 15:27 |
| sdake | and that is where you want the docker dir and host os | 15:28 |
| SamYaple | no | 15:28 |
| SamYaple | why would you care about that being on ssd? | 15:28 |
| SamYaple | putting your os on an ssd is a huge waste of space | 15:28 |
| sdake | it takes 22 minutes to deploy without ssd, 14 mins with ssd | 15:28 |
| SamYaple | and? | 15:28 |
| sdake | that answers why you would want it on an ssd :) | 15:29 |
| SamYaple | 7 minutes? for an initial deploy? doubtful | 15:29 |
| SamYaple | host os on a single disk is a horrible idea | 15:29 |
| sdake | you mean you doubt my measurements or doubt oeple care? | 15:29 |
| SamYaple | i doubt you are considering production architectures at all | 15:29 |
| *** mfalatic has joined #kolla | 15:29 | |
| sdake | could you go into more detail | 15:31 |
| SamYaple | ref arch: 2xssd or 10k drive for host os, 6x 1-2TB disk | 15:32 |
| *** daneyon_ has quit IRC | 15:32 | |
| SamYaple | thats an ideal ceph node for me | 15:32 |
| SamYaple | well +1pcie ssd | 15:32 |
| SamYaple | putting your host os on a SPOF is a bad idea | 15:32 |
| *** daneyon has joined #kolla | 15:34 | |
| sdake | pcie is extremely expensive per mb | 15:36 |
| sdake | i think expecting 3 in a box is unliekly | 15:36 |
| sdake | especially since i'm not even sure if there aare 3 pci in most boxes | 15:36 |
| SamYaple | what are you talking about? | 15:36 |
| SamYaple | oh no | 15:36 |
| SamYaple | 2xssd is sata/scsi junk drives | 15:36 |
| SamYaple | you dont need much perforamnce there | 15:36 |
| SamYaple | anyway the point is no one colocates thier host os on a single pcie | 15:37 |
| SamYaple | as you said its expensive per mb and most of the host os just sits there | 15:37 |
| SamYaple | not to mention being a single point of failure | 15:37 |
| sdake | the model i have is 6 machines out of 100 node cluster with 6-12 disks each plus pcie cache device | 15:40 |
| SamYaple | model based on what? | 15:41 |
| *** diga has joined #kolla | 15:41 | |
| *** daneyon has quit IRC | 15:41 | |
| *** sdake has quit IRC | 15:53 | |
| *** sdake has joined #kolla | 15:53 | |
| sdake | so for the demo i want thinks to be speedy, and if we dont automate it, how would i do it in a non-automated session? | 15:56 |
| sdake | thinks/things | 15:56 |
| SamYaple | by manually doing all the things in the scripts | 15:56 |
| *** blahRus has joined #kolla | 15:56 | |
| sdake | where are the scrpts | 15:57 |
| *** jtriley has joined #kolla | 15:59 | |
| *** mbound has quit IRC | 16:01 | |
| SamYaple | the ceph-osd has the script that does the bootstrap | 16:01 |
| SamYaple | youll have to pick that apart | 16:01 |
| SamYaple | make sure it lines up properly with the partition names in the right locations | 16:01 |
| *** sdake_ has joined #kolla | 16:02 | |
| *** daneyon has joined #kolla | 16:02 | |
| *** sdake has quit IRC | 16:04 | |
| *** jtriley has quit IRC | 16:04 | |
| *** cloudnautique has joined #kolla | 16:06 | |
| *** shardy_ has joined #kolla | 16:08 | |
| *** shardy has quit IRC | 16:10 | |
| *** shardy_ has quit IRC | 16:14 | |
| *** mbound has joined #kolla | 16:14 | |
| *** unicell has quit IRC | 16:14 | |
| *** shardy has joined #kolla | 16:15 | |
| *** mbound has quit IRC | 16:19 | |
| sdake_ | SamYaple looking at hte ceph apatch it looks extrremely involved to modify it to not reformat the disk if it is marked | 16:21 |
| SamYaple | that is what i said | 16:23 |
| sdake_ | i mean to do it manully | 16:25 |
| *** sdake_ is now known as sdake | 16:25 | |
| sdake | isn't that what you said? | 16:25 |
| SamYaple | it is complicated to do it manually | 16:26 |
| SamYaple | i do not recommend it | 16:26 |
| sdake | so demo at summit without a cache layer then? | 16:27 |
| SamYaple | do you seriously not have any spare disks? | 16:28 |
| *** Ti-mo has quit IRC | 16:28 | |
| sdake | i have mini dell machine | 16:28 |
| sdake | they have room for 1 3.5inch disk | 16:28 |
| sdake | and they have 2 pcie slots | 16:28 |
| sdake | so no | 16:28 |
| *** vincent_1dk has quit IRC | 16:29 | |
| *** Windir has quit IRC | 16:29 | |
| *** kelkoobenoitr has quit IRC | 16:29 | |
| sdake | i have 3 of the machines | 16:29 |
| SamYaple | can we use other hardware then? that doesnt sound like a good demo | 16:29 |
| SamYaple | without without the cache I am deploying >14minutes | 16:29 |
| sdake | perhaps it was a bad choice to buy the minidellls at the tie they seemed good price for what i ineeded | 16:29 |
| *** tfukushima has quit IRC | 16:29 | |
| SamYaple | <14minutes | 16:29 |
| sdake | no what I mean is dont use the docker ceph cache | 16:29 |
| SamYaple | less than | 16:29 |
| sdake | the patch you just worked up | 16:29 |
| SamYaple | oh well were demonig liberty right? | 16:30 |
| SamYaple | wait youre trying to piece together something from master for this? | 16:30 |
| sdake | ya, I think its always cool to show in a demo things to come | 16:30 |
| *** vinkman has left #kolla | 16:30 | |
| *** vincent_vdk has joined #kolla | 16:30 | |
| SamYaple | hmm | 16:30 |
| *** tfukushima has joined #kolla | 16:30 | |
| sdake | but I guess its ok | 16:30 |
| *** Ti-mo has joined #kolla | 16:31 | |
| *** bkosick has joined #kolla | 16:33 | |
| *** Windir has joined #kolla | 16:36 | |
| *** dwalsh has quit IRC | 16:42 | |
| *** kelkoobenoitr has joined #kolla | 16:42 | |
| *** tfukushima has quit IRC | 16:43 | |
| *** exploreshaifali has quit IRC | 16:45 | |
| *** bkosick has quit IRC | 16:48 | |
| *** jasonsb has quit IRC | 16:49 | |
| *** sdake has quit IRC | 16:51 | |
| *** unicell has joined #kolla | 16:54 | |
| *** openstackgerrit has quit IRC | 17:01 | |
| *** openstackgerrit has joined #kolla | 17:01 | |
| openstackgerrit | Manoj Kumar Jain proposed openstack/kolla: DOC changes for ALL-IN-ONE Deployment https://review.openstack.org/236667 | 17:02 |
| openstackgerrit | Merged openstack/kolla: Add build profiles to build.py https://review.openstack.org/234229 | 17:02 |
| *** mandre has quit IRC | 17:19 | |
| *** vinkman has joined #kolla | 17:20 | |
| *** mandre has joined #kolla | 17:20 | |
| *** achanda has joined #kolla | 17:29 | |
| *** achanda has quit IRC | 17:29 | |
| *** achanda has joined #kolla | 17:30 | |
| *** dwalsh has joined #kolla | 17:36 | |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Automate setting up a cache tier https://review.openstack.org/236756 | 17:38 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Allow ceph pool names to be configurable https://review.openstack.org/236755 | 17:38 |
| *** jasonsb has joined #kolla | 17:39 | |
| *** vinkman has left #kolla | 17:44 | |
| *** athomas has quit IRC | 17:46 | |
| *** sdake has joined #kolla | 17:46 | |
| *** jasonsb has quit IRC | 17:47 | |
| *** diga has quit IRC | 17:50 | |
| *** sdake has quit IRC | 18:02 | |
| SamYaple | so with the cache setup and untuned i get ~4000 4k iops | 18:06 |
| SamYaple | not bad | 18:06 |
| SamYaple | inside a single instance* | 18:06 |
| SamYaple | i can do that from several instances at once | 18:06 |
| SamYaple | but raw perforamnce is only ~500MB | 18:07 |
| SamYaple | not satuating anymore | 18:07 |
| *** sdake has joined #kolla | 18:11 | |
| *** loth has joined #kolla | 18:12 | |
| *** sdake_ has joined #kolla | 18:15 | |
| sdake_ | yo | 18:15 |
| SamYaple | sdake_: did you see above^^ | 18:16 |
| *** sdake has quit IRC | 18:16 | |
| sdake_ | no i was disconnected because wife was rebooting wireless ithink | 18:16 |
| SamYaple | ah so some numbers | 18:16 |
| SamYaple | with the cache setup and untuned i get ~4000 4k iops inside the instance (instead of 5k) | 18:16 |
| SamYaple | i can do that from several instances at once | 18:16 |
| SamYaple | but raw perforamnce is only ~500MB (instead of saturating 10gb) | 18:17 |
| SamYaple | these numbers make sense and arent bad | 18:17 |
| SamYaple | i would like to see it at scale | 18:17 |
| sdake_ | without cache 1000iops? | 18:18 |
| SamYaple | no without cache its like 100 iops | 18:18 |
| SamYaple | thats to be expected | 18:18 |
| sdake_ | only one cache device? | 18:19 |
| SamYaple | two | 18:19 |
| sdake_ | two 750s? | 18:19 |
| SamYaple | seperate hosts | 18:19 |
| SamYaple | yea | 18:19 |
| sdake_ | i'd like o have a roadmap slide | 18:19 |
| sdake_ | that will be something we can talk about | 18:19 |
| SamYaple | probably need to have a roadmap first | 18:20 |
| sdake_ | roadmap wrong word | 18:20 |
| sdake_ | things already under investigation | 18:20 |
| sdake_ | we have a roadmap sessoin at 3:40 iirc | 18:20 |
| *** devlaps has joined #kolla | 18:22 | |
| sdake_ | http://mitakadesignsummit.sched.org/event/f9ef3969e1b69621415a3b68d08e9e04#.ViU1ChCrTBI | 18:23 |
| SamYaple | asalkeld: was asking about schedule things sdake_ | 18:24 |
| sdake_ | ya i put in half the schedule last night | 18:24 |
| sdake_ | going to do the rest today | 18:24 |
| *** sdake_ is now known as sdake | 18:25 | |
| sdake | so from 100 iops to 4k iops just by adding two cache devices? | 18:25 |
| *** jasonsb has joined #kolla | 18:26 | |
| SamYaple | well sure, but thats only a few osds | 18:26 |
| SamYaple | if i had 100 osds it would be different | 18:26 |
| SamYaple | not _fast_ but different | 18:26 |
| sdake | osd = disk? | 18:26 |
| SamYaple | osd is where ceph stores its objects | 18:28 |
| *** shardy is now known as shardy_afk | 18:30 | |
| sdake | osd is a softwar eprocess then? where there is 1 per physical disk? | 18:30 |
| *** dwalsh_ has joined #kolla | 18:31 | |
| SamYaple | yea thats close enough to the truth | 18:31 |
| sdake | so we start ultiple osd containers then on a single node? | 18:31 |
| SamYaple | potentially | 18:32 |
| sdake | right for example if a mahcine has 20 disks, there would be 20 osd containers on it? | 18:32 |
| SamYaple | correct | 18:32 |
| sdake | if that machine fails is ther eanti affinity of the storage to make suer its replicated to other hosts? | 18:32 |
| SamYaple | the way i set it up is on a host level | 18:33 |
| SamYaple | you can set the crush map however you want | 18:33 |
| SamYaple | osd host rack datacenter etc | 18:33 |
| sdake | if the inventory hs 3 storage nodes, it will replicte to all 3 then? | 18:34 |
| SamYaple | out of the box yes it will need at _least_ 3 different hosts to do that | 18:34 |
| SamYaple | otherwise it will say its unclean | 18:34 |
| sdake | when i ran with 1 osd it cratered | 18:35 |
| SamYaple | you just didnt follow the instructions | 18:35 |
| SamYaple | it didnt crash | 18:35 |
| sdake | oh i see you mean to modify the replication count | 18:35 |
| SamYaple | yes | 18:35 |
| sdake | well it did crater whether or not i followed the dirctions :) | 18:35 |
| sdake | and was unrecoverable | 18:36 |
| sdake | when do you want to get together and do slides | 18:36 |
| SamYaple | dunno. im really tired | 18:37 |
| sdake | not today | 18:37 |
| SamYaple | im back on first shift now | 18:37 |
| sdake | but tomorrow-wed | 18:37 |
| SamYaple | lets do it after the summit | 18:37 |
| sdake | slides? | 18:38 |
| SamYaple | im kidding | 18:38 |
| SamYaple | i make jokes | 18:38 |
| SamYaple | its what i do | 18:38 |
| SamYaple | sdake: try out that clusterer stuff in the queue | 18:40 |
| SamYaple | i need centos confirmation | 18:40 |
| sdake | mostly focused on summit this week sam | 18:40 |
| SamYaple | what happened to wanting to do new things | 18:41 |
| sdake | i'll see if i can get to it but i hae to reload my machines | 18:41 |
| SamYaple | youve been takingabout new ceph stuff for like 3 hours | 18:41 |
| sdake | because i want to demo it :) | 18:41 |
| sdake | but it seems that wont be possible | 18:41 |
| SamYaple | demo cluster cause its awesome | 18:41 |
| sdake | atleast the caching | 18:41 |
| sdake | need to mkae a video of the demo just icase shit goes wrong | 18:43 |
| SamYaple | nah well just log into my boxes. they never fail me | 18:43 |
| sdake | japan | 18:43 |
| SamYaple | ive got tunnels out the wazoo | 18:44 |
| *** jasonsb has quit IRC | 18:50 | |
| *** sdake has quit IRC | 19:00 | |
| *** jasonsb has joined #kolla | 19:03 | |
| *** cloudnautique has quit IRC | 19:10 | |
| *** inc0 has joined #kolla | 19:23 | |
| inc0 | hey | 19:23 |
| inc0 | guys do we have design session schedule somewhere? | 19:23 |
| coolsvap | inc0, https://www.openstack.org/summit/tokyo-2015/schedule/design-summit | 19:27 |
| inc0 | yeah, but most of work sessions aren't planned yet | 19:28 |
| inc0 | I meant which topic will be discussen on which slot | 19:28 |
| coolsvap | i think it is decided, there are arranged as per time slots | 19:32 |
| coolsvap | like https://mitakadesignsummit.sched.org/overview/type/kolla?iframe=no&w=i:100;&sidebar=yes&bg=no | 19:33 |
| inc0 | yeah, but we had etherpad with work session topics prepaired | 19:33 |
| inc0 | and I wonder which sessions I'm going to miss | 19:33 |
| inc0 | since I have conflicts :S | 19:33 |
| inc0 | what's sad is that one of sessions that conflicts with us is "infrastructure containers | 19:36 |
| inc0 | " | 19:36 |
| *** exploreshaifali has joined #kolla | 19:50 | |
| *** inc0 has quit IRC | 19:55 | |
| *** cemason has quit IRC | 20:04 | |
| *** cemason has joined #kolla | 20:05 | |
| *** jtriley has joined #kolla | 20:10 | |
| *** cloudnautique has joined #kolla | 20:14 | |
| *** jtriley has quit IRC | 20:15 | |
| *** dims_ has joined #kolla | 20:16 | |
| *** diogogmt has quit IRC | 20:16 | |
| *** dims has quit IRC | 20:19 | |
| *** jtriley has joined #kolla | 20:19 | |
| *** achanda has quit IRC | 20:26 | |
| openstackgerrit | Merged openstack/kolla: Use bool filter for ironic https://review.openstack.org/236733 | 20:27 |
| openstackgerrit | Merged openstack/kolla: Update requirements to use tarball by default https://review.openstack.org/236914 | 20:29 |
| *** gfidente has quit IRC | 20:43 | |
| *** achanda has joined #kolla | 20:46 | |
| *** jtriley has quit IRC | 20:47 | |
| *** jtriley has joined #kolla | 20:53 | |
| *** tpot has joined #kolla | 20:58 | |
| *** dwalsh_ has quit IRC | 21:26 | |
| *** dwalsh has quit IRC | 21:26 | |
| *** jtriley has quit IRC | 21:27 | |
| *** tpot has quit IRC | 21:29 | |
| *** achanda has quit IRC | 21:34 | |
| Hui | the latest neutron-server broken | 21:39 |
| Hui | has anyone observed this | 21:39 |
| Hui | ERROR neutron ProgrammingError: (_mysql_exceptions.ProgrammingError) (1146, "Table 'neutron.ml2_vlan_allocations' doesn't exist") [SQL: u'SELECT ml2_vlan_allocations.physical_network AS ml2_vlan_allocations_physical_network, ml2_vlan_allocations.vlan_id AS ml2_vlan_allocations_vlan_id, ml2_vlan_allocations.allocated AS ml2_vlan_allocations_allocated \nFROM ml2_vlan_allocations FOR UPDATE' | 21:40 |
| *** achanda has joined #kolla | 21:50 | |
| *** daneyon has quit IRC | 22:08 | |
| *** exploreshaifali has quit IRC | 22:14 | |
| *** dims_ has quit IRC | 22:15 | |
| *** cemason has quit IRC | 22:32 | |
| *** cemason has joined #kolla | 22:33 | |
| *** loth has left #kolla | 22:36 | |
| *** achanda has quit IRC | 22:55 | |
| *** achanda_ has joined #kolla | 22:55 | |
| *** achanda has joined #kolla | 22:55 | |
| *** achanda_ has quit IRC | 22:59 | |
| *** sdake has joined #kolla | 23:04 | |
| *** blahRus has quit IRC | 23:18 | |
| sdake | "come with me, if you want to live." | 23:33 |
| *** cemason has quit IRC | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!