| *** akapil has joined #openstack-manila | 00:02 | |
| *** akapil has quit IRC | 00:06 | |
| *** kaisers has joined #openstack-manila | 00:13 | |
| *** kaisers has quit IRC | 00:17 | |
| *** cknight has quit IRC | 00:19 | |
| *** vbellur has joined #openstack-manila | 01:05 | |
| *** kberger has joined #openstack-manila | 01:09 | |
| openstackgerrit | Tuan Luong-Anh proposed openstack/manila: Avoid Forcing the Translation of Translatable Variables https://review.openstack.org/378446 | 01:10 |
|---|---|---|
| *** tuanluong has joined #openstack-manila | 01:55 | |
| *** kaisers has joined #openstack-manila | 02:02 | |
| *** kaisers has quit IRC | 02:06 | |
| *** gouthamr has quit IRC | 02:07 | |
| *** winston-d_ has joined #openstack-manila | 02:29 | |
| *** akapil has joined #openstack-manila | 02:51 | |
| openstackgerrit | Merged openstack/manila: Update the home page https://review.openstack.org/276680 | 02:56 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/manila: Updated from global requirements https://review.openstack.org/380095 | 03:27 |
| *** akapil has quit IRC | 03:41 | |
| *** akapil has joined #openstack-manila | 03:43 | |
| *** kaisers has joined #openstack-manila | 03:50 | |
| *** kaisers has quit IRC | 03:55 | |
| *** dsariel_ has quit IRC | 03:59 | |
| *** mtanino has quit IRC | 04:03 | |
| *** akapil has quit IRC | 04:49 | |
| *** winston-d_ has quit IRC | 04:56 | |
| *** narayrak has joined #openstack-manila | 04:56 | |
| *** akapil has joined #openstack-manila | 04:58 | |
| *** narayrak has quit IRC | 04:59 | |
| *** akapil has quit IRC | 05:00 | |
| *** akapil has joined #openstack-manila | 05:01 | |
| *** akapil_ has joined #openstack-manila | 05:09 | |
| *** akapil has quit IRC | 05:13 | |
| *** shausy has joined #openstack-manila | 05:28 | |
| *** akapil_ has quit IRC | 05:28 | |
| *** kaisers has joined #openstack-manila | 05:39 | |
| openstackgerrit | Bhagyashri Shewale proposed openstack/python-manilaclient: Replace functions 'Dict.get' and 'del' with 'Dict.pop' https://review.openstack.org/361858 | 05:40 |
| *** dsariel_ has joined #openstack-manila | 05:43 | |
| *** kaisers has quit IRC | 05:48 | |
| *** kaisers has joined #openstack-manila | 05:55 | |
| *** nkrinner_afk is now known as nkrinner | 06:04 | |
| *** rraja has joined #openstack-manila | 06:11 | |
| *** rraja has quit IRC | 06:16 | |
| *** adrianofr_ has quit IRC | 06:18 | |
| *** lpetrut has joined #openstack-manila | 06:22 | |
| *** adrianofr_ has joined #openstack-manila | 06:22 | |
| *** rraja has joined #openstack-manila | 06:27 | |
| *** lpetrut has quit IRC | 06:51 | |
| *** adrianofr_ has quit IRC | 06:51 | |
| *** adrianofr_ has joined #openstack-manila | 06:52 | |
| *** lpetrut has joined #openstack-manila | 07:07 | |
| *** lpetrut has quit IRC | 07:21 | |
| *** pcaruana has joined #openstack-manila | 07:26 | |
| *** lpetrut has joined #openstack-manila | 08:31 | |
| *** vponomaryov1 has joined #openstack-manila | 08:38 | |
| *** ociuhandu has joined #openstack-manila | 08:56 | |
| openstackgerrit | Merged openstack/manila: Updated from global requirements https://review.openstack.org/380095 | 09:04 |
| *** ociuhandu has quit IRC | 09:31 | |
| *** kaisers_ has joined #openstack-manila | 09:54 | |
| *** kaisers_ has quit IRC | 09:55 | |
| *** kaisers has quit IRC | 09:58 | |
| *** kaisers has joined #openstack-manila | 09:58 | |
| *** kaisers has quit IRC | 10:00 | |
| *** ociuhandu has joined #openstack-manila | 10:04 | |
| *** kaisers has joined #openstack-manila | 10:10 | |
| *** kaisers has quit IRC | 10:14 | |
| *** kaisers has joined #openstack-manila | 10:27 | |
| *** kaisers has quit IRC | 10:39 | |
| *** kaisers has joined #openstack-manila | 10:40 | |
| *** dsariel_ has quit IRC | 10:46 | |
| *** kaisers has quit IRC | 10:56 | |
| *** tries_ has quit IRC | 11:15 | |
| *** alyson_ has joined #openstack-manila | 11:15 | |
| *** tries_ has joined #openstack-manila | 11:26 | |
| *** nkrinner has quit IRC | 11:34 | |
| *** nkrinner has joined #openstack-manila | 11:46 | |
| *** tpsilva has joined #openstack-manila | 11:48 | |
| *** gcb has joined #openstack-manila | 11:49 | |
| *** ociuhandu has quit IRC | 11:52 | |
| *** gouthamr has joined #openstack-manila | 12:03 | |
| *** tuanluong has quit IRC | 12:04 | |
| openstackgerrit | Valeriy Ponomaryov proposed openstack/python-manilaclient: Fix share-server-delete command https://review.openstack.org/381231 | 12:06 |
| *** dsariel_ has joined #openstack-manila | 12:21 | |
| openstackgerrit | Valeriy Ponomaryov proposed openstack/python-manilaclient: Run pre_test_hook.sh of manila in client's CI https://review.openstack.org/382519 | 12:22 |
| *** kaisers has joined #openstack-manila | 12:24 | |
| *** alyson_ has quit IRC | 12:36 | |
| *** adrianofr_ has quit IRC | 12:37 | |
| *** alyson_ has joined #openstack-manila | 12:40 | |
| *** adrianofr_ has joined #openstack-manila | 12:40 | |
| *** kaisers has quit IRC | 12:51 | |
| *** ociuhandu has joined #openstack-manila | 12:56 | |
| openstackgerrit | Valeriy Ponomaryov proposed openstack/python-manilaclient: Add support of Identity API V3 in functional tests https://review.openstack.org/340739 | 12:57 |
| *** akerr has joined #openstack-manila | 12:57 | |
| *** ociuhandu has quit IRC | 12:59 | |
| *** ociuhandu has joined #openstack-manila | 13:04 | |
| *** shausy has quit IRC | 13:08 | |
| *** gcb has quit IRC | 13:15 | |
| openstackgerrit | Valeriy Ponomaryov proposed openstack/python-manilaclient: Add support of Identity API V3 in functional tests https://review.openstack.org/340739 | 13:26 |
| *** mkoderer has joined #openstack-manila | 13:27 | |
| openstackgerrit | Valeriy Ponomaryov proposed openstack/python-manilaclient: Add support of Identity API V3 in functional tests https://review.openstack.org/340739 | 13:28 |
| *** dustins has joined #openstack-manila | 13:28 | |
| *** dustins has quit IRC | 13:33 | |
| *** porrua has joined #openstack-manila | 13:38 | |
| *** merooney has joined #openstack-manila | 13:46 | |
| *** dustins has joined #openstack-manila | 13:47 | |
| *** eharney has joined #openstack-manila | 13:47 | |
| *** akshai has joined #openstack-manila | 13:49 | |
| *** cknight has joined #openstack-manila | 13:49 | |
| *** akshai_ has joined #openstack-manila | 13:50 | |
| *** rooneym has joined #openstack-manila | 13:51 | |
| *** ChanServ changes topic to "OpenStack Shared File Systems | Manila| Newton is released!" | 13:51 | |
| *** akshai has quit IRC | 13:54 | |
| *** mtanino has joined #openstack-manila | 13:56 | |
| *** kaisers has joined #openstack-manila | 14:14 | |
| *** xyang_ has joined #openstack-manila | 14:17 | |
| *** jcsp has joined #openstack-manila | 14:27 | |
| *** gouthamr has quit IRC | 14:29 | |
| *** gouthamr has joined #openstack-manila | 14:31 | |
| *** akapil has joined #openstack-manila | 14:46 | |
| *** kaisers has quit IRC | 14:47 | |
| *** lpetrut has quit IRC | 14:49 | |
| *** ganso has joined #openstack-manila | 14:50 | |
| *** akapil has quit IRC | 14:51 | |
| *** erlon has quit IRC | 14:54 | |
| *** DuncanT has quit IRC | 14:55 | |
| *** mkoderer has quit IRC | 14:55 | |
| *** cargonza has quit IRC | 14:55 | |
| *** markstur_ has joined #openstack-manila | 14:56 | |
| *** kaisers has joined #openstack-manila | 14:59 | |
| *** timcl1 has joined #openstack-manila | 14:59 | |
| *** timcl has quit IRC | 14:59 | |
| *** markstur__ has joined #openstack-manila | 15:00 | |
| *** markstur___ has joined #openstack-manila | 15:01 | |
| *** markstur_ has quit IRC | 15:03 | |
| *** markstur__ has quit IRC | 15:04 | |
| *** mkoderer has joined #openstack-manila | 15:21 | |
| *** merooney has quit IRC | 15:29 | |
| *** erlon has joined #openstack-manila | 15:36 | |
| *** DuncanT has joined #openstack-manila | 15:36 | |
| *** markstur_ has joined #openstack-manila | 15:41 | |
| *** cargonza has joined #openstack-manila | 15:41 | |
| *** markstur___ has quit IRC | 15:45 | |
| *** kaisers has quit IRC | 15:53 | |
| *** akshai_ has quit IRC | 15:55 | |
| rraja | bswartz: hi! I thought the only missing part of ganesha, was that it was missing dynamic update of exports and that it's been added in the latest ganesha release. | 16:01 |
| rraja | bswartz: what are the other missing essential ganesha features? | 16:01 |
| tbarron | rraja: +++ | 16:02 |
| vponomaryov1 | rraja: "Modify share permissions" | 16:02 |
| vponomaryov1 | rraja: not reloading/restarting anything | 16:02 |
| vponomaryov1 | rraja: is it addressed too? | 16:03 |
| *** surabujin_ has quit IRC | 16:03 | |
| *** kberger has quit IRC | 16:05 | |
| *** kberger has joined #openstack-manila | 16:05 | |
| *** nkrinner is now known as nkrinner_afk | 16:09 | |
| *** markstur__ has joined #openstack-manila | 16:09 | |
| rraja | vponomaryov1: that's what I assumed these commits did https://github.com/nfs-ganesha/nfs-ganesha/commits/2f47e8a761f3700 | 16:10 |
| rraja | vponomaryov1: as in dynamically modify an export block without restarting ganesha server | 16:11 |
| *** akshai has joined #openstack-manila | 16:11 | |
| vponomaryov1 | rraja: so, version 2.4+ should work for us, do I understand you correctly? | 16:12 |
| *** markstur_ has quit IRC | 16:12 | |
| rraja | vponomaryov1: yep. I wrote about this to manila devel, here http://osdir.com/ml/openstack-dev/2016-08/msg01211.html | 16:14 |
| rraja | vponomaryov1: unless ofc I totally misunderstood frank filz | 16:14 |
| vponomaryov1 | rraja: good news, thanks | 16:19 |
| rraja | vponomaryov1: are there other features that you know of that makes ganesha unusable? | 16:21 |
| rraja | vponomaryov1: we need to communicate that to ganesha community as quickly as we can. | 16:22 |
| vponomaryov1 | rraja: nothing comes to my mind now | 16:22 |
| *** surabujin_ has joined #openstack-manila | 16:24 | |
| rraja | vponomaryov1: thanks valeriy. if we find bugs i'm sure they'd try to help us. | 16:24 |
| *** jcsp has quit IRC | 16:25 | |
| vponomaryov1 | rraja: I guess you'd bettwer weigh in https://etherpad.openstack.org/p/manila-ocata-design-summit-topics | 16:25 |
| vponomaryov1 | rraja: in appropriate topic | 16:26 |
| rraja | vponomaryov1: ok. will do. | 16:26 |
| tbarron | vponomaryov1: I put a topic there for this and put rraja's name next to it in addition to mine | 16:26 |
| tbarron | bswartz: vponomaryov1 yeah, b/c of the mail rraja sent earlier I was quite surprised to hear there are outstanding issues with ganesha | 16:27 |
| vponomaryov1 | rraja, tbarron: lines 45-47 | 16:27 |
| bswartz | tbarron: you back from meeting yet? | 16:27 |
| tbarron | so we need to find what other issues there are | 16:27 |
| tbarron | bswartz: not really, just couldn't resist | 16:28 |
| bswartz | tbarron: main issues with ganesha are: you can't modify the permissions on an existing share, and you can't add new shares without bouncing the whole service | 16:28 |
| vponomaryov1 | bswartz: looks line not anymore | 16:29 |
| bswartz | vponomaryov1: what? | 16:29 |
| vponomaryov1 | see above | 16:29 |
| bswartz | ty | 16:29 |
| *** xyang1 has joined #openstack-manila | 16:32 | |
| bswartz | okay so the remaining issue is about whether ganesha can pick up new mounts on the fly | 16:32 |
| bswartz | iirc there was a problem where if you started ganesha first, then mounted a new fileseystem, ganesha couldn't see it | 16:33 |
| *** timcl1 has quit IRC | 16:33 | |
| bswartz | vponomaryov1: does that sound familiar | 16:33 |
| vponomaryov1 | bswartz: personally, I didn't test "pick up new mount of nly" thing, and such case is not mentioned by rraja too | 16:35 |
| *** timcl has joined #openstack-manila | 16:36 | |
| vponomaryov1 | bswartz: so, it is open question for me | 16:36 |
| bswartz | k it may be worth investigating | 16:36 |
| vponomaryov1 | bswartz: anyway, I guess if it is so, we can report such bug to ganesha team and they will fix it | 16:36 |
| bswartz | it's good news for both the existing container driver and also if we refactor it | 16:36 |
| vponomaryov1 | bswartz: also, we can avoid LV problem for it just starting using ZFS for container driver | 16:37 |
| vponomaryov1 | and its ZVOLs | 16:37 |
| bswartz | lol you mentioned that before | 16:37 |
| vponomaryov1 | I know ) | 16:37 |
| bswartz | IIRC not all of the ZFS kernel bits have been make namespace-aware and I expect problems if it's used from inside a container | 16:38 |
| *** merooney has joined #openstack-manila | 16:38 | |
| rraja | bswartz: yep! https://www.mail-archive.com/openstack-dev@lists.openstack.org/msg90381.html :) | 16:38 |
| vponomaryov1 | bswartz: good point, I didn't use it in containers | 16:38 |
| vponomaryov1 | only system-wide | 16:38 |
| bswartz | vponomaryov1: it might work but it wouldn't address any of our serious problems with the interactions between the NFS daemon and the filesystem | 16:39 |
| bswartz | even ZFS has to perform mount operations on new zvols | 16:39 |
| *** akerr has quit IRC | 16:39 | |
| *** akerr has joined #openstack-manila | 16:40 | |
| bswartz | rraja: thanks! | 16:41 |
| bswartz | rraja: are you aware of the other issue I'm talking about? it needs to be confirmed, but I thought that ganesha had an issue creating exports on filesystems which were mounted after the service started up | 16:42 |
| rraja | bswartz: no I wasn't until now. i'll do that. thanks! | 16:42 |
| *** dsariel_ has quit IRC | 16:43 | |
| bswartz | rraja: it's not a confirmed problem -- we need to test that case | 16:43 |
| bswartz | it's just something I remember aovchinnikov mentioning | 16:43 |
| bswartz | tbarron: it will be lunch time soon here but ping me when you're back | 16:44 |
| rraja | bswartz: i'll test that too. thanks! | 16:48 |
| tbarron | bswartz: will do. | 16:48 |
| vponomaryov1 | can I ask some kind core people to look at this change -https://review.openstack.org/#/c/382519/ in manilalicent? It is expected to make our unstable manilaclient functional tests job more stable. | 16:51 |
| *** markstur_ has joined #openstack-manila | 16:58 | |
| *** markstur__ has quit IRC | 17:01 | |
| *** markstur__ has joined #openstack-manila | 17:02 | |
| *** markstur__ has quit IRC | 17:04 | |
| *** markstur_ has quit IRC | 17:04 | |
| *** lpetrut has joined #openstack-manila | 17:43 | |
| openstackgerrit | Merged openstack/python-manilaclient: Run pre_test_hook.sh of manila in client's CI https://review.openstack.org/382519 | 17:47 |
| bswartz | tbarron: back | 17:48 |
| tbarron | bswartz: hi | 17:49 |
| bswartz | tbarron: done w/ meeting? | 17:49 |
| tbarron | yes, even ate a sandwich too | 17:49 |
| bswartz | cool | 17:50 |
| bswartz | so on the topic of the service image | 17:50 |
| bswartz | our stance has always been that distros should provide their own manila generic driver compatible service images | 17:50 |
| bswartz | IMO we have failed to make that an easy thing to do | 17:51 |
| tbarron | i think that's a reasonable expectation | 17:51 |
| bswartz | well that stance is based on the fact that it would be impossible for us to deal with all of the distro-specifics upstream | 17:51 |
| tbarron | (and we probably have to sign the image, etc. if we ship it anyways) | 17:51 |
| tbarron | vkmc ^^^^ | 17:52 |
| bswartz | so the problem is that the reference image is based on Ubuntu, and we haven't done enough to document the requirements so that "real" images could be produced | 17:52 |
| tbarron | she's probably busy now but has experience with DIB from working on trove | 17:52 |
| bswartz | if actual (production) users are using the reference image, then I feel we've failed | 17:53 |
| bswartz | so what I'd like to see done here is a spec for all of the interactions between the generic driver and the service VM | 17:53 |
| bswartz | and I'd like to see multiple working implementations, beyond the reference one | 17:54 |
| tbarron | rraja you too ^^^ | 17:54 |
| tbarron | we want to gain a better understanding of the generic driver, its image requirements, what makes it fragile | 17:55 |
| bswartz | tbarron: sneak preview, I'm working on a new service image to address the goals I laid out | 17:55 |
| bswartz | tbarron: fragile? | 17:55 |
| openstackgerrit | Valeriy Ponomaryov proposed openstack/python-manilaclient: Add support of Identity API V3 in functional tests https://review.openstack.org/340739 | 17:56 |
| rraja | bswartz: cool! | 17:56 |
| tbarron | bswartz: it *looks* like it is because of CI; so part of what we want to get to is an understanding of how to do generic in CI reliably | 17:56 |
| bswartz | tbarron: correct | 17:56 |
| bswartz | the generic driver itself is fine, and so is the reference image | 17:57 |
| bswartz | the problem is that it's resource hungry, and the gate is resource starved | 17:57 |
| tbarron | bswartz: so right now I would say we don't *know* empirically one way or the other, though you may have good reasons for what you are saying | 17:57 |
| bswartz | you combine there 2 facts and the result is lots of engineer tears | 17:57 |
| tbarron | well, that is clearly *a* problem, perhaps the only one, but we don't to my knowledge have testing without those conditions so we don't know for sure | 17:58 |
| bswartz | tbarron: I run the generic driver on my laptop all the time -- never had problems with instability -- just resource exhaustion | 17:58 |
| tbarron | not trying to be contentious though | 17:58 |
| tbarron | k, i'm willing on that basis to adopt your assertions as working hypothesis ... | 17:59 |
| vponomaryov1 | tbarron: how to do CI reliable using generic driver? - simple, just do reliable Nova, Cinder, Neutron and Glance, that's all | 18:00 |
| tbarron | that said, we want to look at adapting the generic driver to be NFS gateway for other backends (e.g. cephfs) working at cloud rather than laptop scale | 18:00 |
| tbarron | vponomaryov1: so you are making an argument against bswartz | 18:00 |
| vponomaryov1 | markstur, tbarron: thatnks for merge of manilaclient change! | 18:00 |
| vponomaryov1 | tbarron: why against? | 18:00 |
| vponomaryov1 | tbarron: I am saying the same in other words | 18:00 |
| tbarron | he is saying the only fundamental issues are resource constraints whereas you are arguiing that the external dependencies are a significant issue | 18:01 |
| bswartz | vponomaryov1: we should stop blaming those other projects for our bugs | 18:01 |
| vponomaryov1 | tbarron: its dependencies are unstable, that is why CI is unstable | 18:01 |
| bswartz | vponomaryov1: at least we should assume the bugs are our until proven otherwise | 18:01 |
| bswartz | s/our/ours/ | 18:01 |
| *** akapil has joined #openstack-manila | 18:01 | |
| tbarron | I have myself made the vponomaryov1 argument: that generic driver is going to be inherently fragile irrespective of resources b/c of all its external dependencies. | 18:02 |
| vponomaryov1 | bswartz: CI instability is caused not by driver bugs | 18:02 |
| vponomaryov1 | bswartz: i Do not say it does not have bugs | 18:02 |
| tbarron | but right now i'm stepping back and am trying to figure it out, because we are looking at using it or something like it for productiion. | 18:03 |
| bswartz | well for example the pre test hook bug in manilaclient CI | 18:03 |
| bswartz | it's not a driver bug but it's still _our_ bug | 18:03 |
| vponomaryov1 | bswartz: yes, agree | 18:04 |
| *** markstur_ has joined #openstack-manila | 18:04 | |
| *** markstur_ has left #openstack-manila | 18:04 | |
| tbarron | and we'd actually like to do HA with service VMs, adding another layer of potential complexity | 18:04 |
| bswartz | tbarron: I'd like to understand that more clearly | 18:04 |
| vponomaryov1 | tbarron: also, don't forget that generic driver works only with openvswitch | 18:04 |
| bswartz | tbarron: we can eliminate some single points of failure but eliminating ALL of them will require serious engineering | 18:05 |
| vponomaryov1 | tbarron: we have support of linuxbridges, but it is not tested at all | 18:05 |
| vponomaryov1 | tbarron: and very likely to be broken | 18:05 |
| *** lpetrut has quit IRC | 18:05 | |
| *** lpetrut has joined #openstack-manila | 18:05 | |
| tbarron | bswartz: vponomaryov1 so you are bringing up good points, and I've been arguing that this would take multiple releases :) | 18:06 |
| vponomaryov1 | tbarron: what "this"? | 18:06 |
| tbarron | bswartz: vponomaryov1 redundant nfs gateways using service images, suitable for production use | 18:07 |
| tbarron | ^^^^ "this" | 18:07 |
| *** akapil_ has joined #openstack-manila | 18:08 | |
| bswartz | tbarron: the redundant gateways thing is a challenge | 18:08 |
| *** markstur has left #openstack-manila | 18:08 | |
| tbarron | "this" would presumably be an evolution of the generic driver, or maybe it could be done with container namespaces on multiple "container" nodes | 18:08 |
| bswartz | Linux NFS server is not meant to be used in that way | 18:08 |
| tbarron | bswartz: agree, just tipping my hand | 18:09 |
| bswartz | in fact I'm not aware of ANY open source NFS daemon which supports clustering | 18:09 |
| vponomaryov1 | tbarron: it would require multi-attach feature from CInder | 18:09 |
| bswartz | vponomaryov1: yes that too, but that is coming | 18:10 |
| vponomaryov1 | tbarron: or HA of volumes too? | 18:10 |
| tbarron | so, back to basics, I tend to agree with bswartz that the newbies to this area (me, rraja, vkmc, ) should start by joining you in thinking about how to create service images | 18:10 |
| tbarron | and learning the requirements on them today | 18:10 |
| *** akapil has quit IRC | 18:11 | |
| tbarron | with the aim of understanding how light weight we can get generic for CI | 18:11 |
| bswartz | tbarron: we need to write down the set of assumptions the generic driver makes about the service image, and what the specific requirements are | 18:11 |
| vponomaryov1 | http://docs.openstack.org/developer/manila/devref/generic_driver.html#requirements-for-service-image ? | 18:11 |
| vkmc | o/ | 18:11 |
| bswartz | vponomaryov1: oh yes I forgot about this | 18:11 |
| tbarron | vponomaryov1: sorry, i wasn't clear, i was talking about backing a child-of-generic with cephfs | 18:12 |
| bswartz | vponomaryov1: it's a good start but it needs more detail | 18:12 |
| tbarron | vkmc: we are talking about image buildingn so i just wanted you to know :) | 18:12 |
| vkmc | yeah, catching up with backlog :) | 18:12 |
| bswartz | tbarron: ceph with blocks behind cinder? | 18:12 |
| vkmc | something I noticed is that you mentioned specifically the generic driver | 18:13 |
| rraja | vponomaryov1: bswartz: as in have the service VM talk to CephFS. | 18:13 |
| vkmc | and for what I saw in https://github.com/openstack/manila-image-elements | 18:13 |
| vkmc | there are elements for some of the drivers as well | 18:13 |
| tbarron | bswartz: cephfs, not cinder/blocks | 18:14 |
| bswartz | rraja tbarron: that's a bad fit for the generic driver -- the generic driver is really built around cinder | 18:14 |
| *** markstur has joined #openstack-manila | 18:14 | |
| vponomaryov1 | tbarron, then woudn't it be better to use container driver as layout? | 18:14 |
| bswartz | it sounds like you want a new driver which shares a large amount of generic driver infrastructure | 18:14 |
| tbarron | bswartz: ++ | 18:14 |
| tbarron | vponomaryov1: maybe | 18:15 |
| bswartz | fortunately generic driver has been structured as pretty modular | 18:15 |
| tbarron | bswartz: vponomaryov1 we want to learn from what you folks have already done and will help improve generic and or container along the way | 18:15 |
| tbarron | we also have been planning to use ganesha -- hence rraja's activism on that front w.r.t. the issues that we were aware of | 18:17 |
| bswartz | tbarron: if you're interested in ganesha then containers are probably a better foundation to build on | 18:17 |
| *** xyang_ has quit IRC | 18:18 | |
| vponomaryov1 | tbarron, then it is definitely the way out - use container driver as layout and then plug there storage-based classes for cephfs or glusterfs in addition to existing LVM | 18:18 |
| bswartz | my interest in generic driver comes from my frustration with ganesha | 18:18 |
| tbarron | bswartz: how much of that frustration is addressed by https://www.mail-archive.com/openstack-dev@lists.openstack.org/msg90381.html ? | 18:19 |
| rraja | vponomaryov1: bswartz : our idea is to use the service_instance module and use service VM images with Ganesha to talk to a Ceph storage cluster. | 18:21 |
| bswartz | tbarron: that's a big component, but that fix isn't part of any supported release yet | 18:21 |
| rraja | vponomaryov1: bswartz : are you saying it's inherently a bad idea? | 18:21 |
| *** mkoderer has quit IRC | 18:22 | |
| bswartz | rraja: I'm glad that our issues are getting addressed -- it's just that we have to live on the bleeding edges to receive the fixes we need | 18:24 |
| bswartz | with nfs-kernel-server this stuff has worked fine for years | 18:24 |
| bswartz | it's a more stable foundation to build on | 18:24 |
| bswartz | I think it's worth doing the container research, and worth experimenting with ganesha so we can report issues we find | 18:25 |
| bswartz | however I could not recommend to a customer to use it until it's been stable for awhile | 18:25 |
| rraja | bswartz: yeah I understand that. in fact nfs-ganesha showcased Manila as one of it's use cases in a presentation a while ago. | 18:26 |
| rraja | s/it's/its/ | 18:26 |
| *** markstur has left #openstack-manila | 18:26 | |
| bswartz | that's awesome | 18:26 |
| rraja | bswartz: slide #18 of http://events.linuxfoundation.org/sites/events/files/slides/Collab14_nfsGanesha.pdf . back in 2014. :) | 18:27 |
| *** alyson_ has quit IRC | 18:29 | |
| vponomaryov1 | rraja: blue aliens with yellow mouthes page? | 18:29 |
| rraja | bswartz: vponomaryov1 : I just wanted to know if it's OK to try implementing a driver along the lines of generic that'd use gateway ganesha VM to talk to an external storage backend. | 18:30 |
| rraja | vponomaryov1: yeah. :) I think Manila has come a long way since then. | 18:30 |
| tbarron | vponomaryov1: my kind of people | 18:31 |
| vponomaryov1 | rraja: sure it is ok to extend those both drivers | 18:31 |
| rraja | vponomaryov1: but you see the same stumbling block of an unstable CI unless of course we have a external powerful CI reporting to Jenkins. | 18:32 |
| *** akapil_ has quit IRC | 18:32 | |
| vponomaryov1 | rraja: it is a little bit annoying, but not a showstopper as you can see | 18:32 |
| vponomaryov1 | rraja: it has always been voting )) | 18:33 |
| tbarron | ideally a 100% open-source driver wouldn't have to run as external 3rd party, but resource requirements might force it | 18:33 |
| bswartz | rraja: it would be an open source driver right? | 18:33 |
| openstackgerrit | boden proposed openstack/manila: Replace retrying with tenacity https://review.openstack.org/380552 | 18:33 |
| rraja | vponomaryov1: manila community's general stance has been that the only true multi-tenant architecture is the network separated gateway model used by the generic driver. | 18:34 |
| rraja | so we thought we could piggy back on that | 18:34 |
| bswartz | rraja: not true | 18:34 |
| rraja | bswartz: :( | 18:35 |
| bswartz | rraja: the container driver was written to also support true multi tenancy | 18:35 |
| tbarron | is the network separated gateway model different though? | 18:36 |
| tbarron | model, not implementation | 18:36 |
| rraja | bswartz: oops! i've been assuming that the model is the same. | 18:36 |
| bswartz | rraja: the container driver has very little in common with the generic driver | 18:38 |
| bswartz | other than the goals they're trying to achieve | 18:38 |
| rraja | bswartz: understood. | 18:39 |
| tbarron | bswartz: apart from the ganesha issues, which we may be able to address, are there other factors driving you back to the generic driver instead of the container driver? | 18:41 |
| bswartz | tbarron: yes, docker. | 18:41 |
| bswartz | docker doesn't allow adding new storage devices to a running container | 18:42 |
| tbarron | docker has lots of churn I hear | 18:42 |
| bswartz | it can be done with tricks | 18:42 |
| bswartz | but that's lame | 18:42 |
| tbarron | flocker? | 18:43 |
| bswartz | IMO that doesn't solve anything | 18:43 |
| * tbarron doesn't know much about this stuff | 18:43 | |
| * tbarron vaguely recalls seeing a flocker to docker backed by cDOT preso once | 18:44 | |
| bswartz | I didn't either, until I got in deep with aovchinnikov on the container driver | 18:44 |
| bswartz | tbarron: the problem is that docker assumes all the storage is setup at container boot time | 18:44 |
| bswartz | once the container is booted, why would you ever way to add/remove storage | 18:44 |
| * bswartz headdesk | 18:44 | |
| bswartz | ever want* to | 18:45 |
| tbarron | well, that's kinda the idea of containers for some folks (kaff, kaff). You want to change something, add a new version of the container. | 18:45 |
| bswartz | tbarron: exactly | 18:45 |
| bswartz | manila's use case is just completely foreign to them | 18:46 |
| bswartz | therefore I have little hope of improvement in that area | 18:46 |
| openstackgerrit | Andreas Jaeger proposed openstack/puppet-manila: Enable release notes translation https://review.openstack.org/383344 | 18:46 |
| *** ociuhandu has quit IRC | 18:46 | |
| *** xyang_ has joined #openstack-manila | 18:47 | |
| *** diablo_rojo has joined #openstack-manila | 18:47 | |
| tbarron | bswartz: now that limitation applies to lvm and to storage that presents itself as a block device but not necessarily to network based storage, right? | 18:48 |
| *** harlowja has quit IRC | 18:50 | |
| openstackgerrit | Andreas Jaeger proposed openstack/python-manilaclient: Enable release notes translation https://review.openstack.org/383376 | 18:50 |
| tpsilva | hey, can I get a help with the lvm driver with cifs? | 18:51 |
| tpsilva | do I need to setup a password for my users somehow? | 18:51 |
| vponomaryov1 | tpsilva: LVM driver uses host's CIFS users | 18:53 |
| vponomaryov1 | tpsilva: so, it is out-of-lvm driver question, just common question for usage of CIFS users | 18:54 |
| rraja | tbarron: ping | 18:54 |
| tbarron | rraja: pong | 18:54 |
| *** xyang_ has quit IRC | 18:54 | |
| tpsilva | vponomaryov1: yes :) | 18:54 |
| rraja | tbarron: it's generally not recommended for two manila-share service instances to talk to the same Ceph storage cluster, right? | 18:55 |
| tpsilva | vponomaryov1: but how can I setup a password for a cifs user? I tried smbpasswd but it does not appear to be like that | 18:55 |
| rraja | tbarron: there was a question on #ceph about it. | 18:55 |
| *** xyang_ has joined #openstack-manila | 18:56 | |
| vponomaryov1 | tpsilva: why? | 18:59 |
| tpsilva | vponomaryov1: I changed the password for the user ubuntu with smbpassword, then I tried to mount it | 19:00 |
| tpsilva | vponomaryov1: mounting with: sudo mount -t cifs //172.24.47.59/share-2294b903-6341-4f7d-b781-1c0af8661910 ~/cifs -o username=ubuntu | 19:01 |
| tpsilva | vponomaryov1: after inputting the password, it says No such device or address | 19:01 |
| tpsilva | vponomaryov1: but if I input a wrong password, it says permission denied | 19:02 |
| tbarron | rraja: so they want to define two native cephfs backends and point to the same ceph cluster with both of them? | 19:02 |
| rraja | tbarron: yes | 19:02 |
| tbarron | rraja: do you know why? I think we've assumed that when people would have multiple cephfs backends they would point them at different clusters. I can certainly understand *that* use case. | 19:03 |
| rraja | tbarron: they were getting HA. what if one manila-share instance went down? | 19:03 |
| rraja | s/getting/getting at/ | 19:04 |
| vponomaryov1 | tpsilva: then why do you think it is password problem? | 19:05 |
| vponomaryov1 | tpsilva: password check is first step | 19:05 |
| *** ociuhandu has joined #openstack-manila | 19:06 | |
| tbarron | rraja: so are they running separate share services on different physical nodes pointing at the same cluster with the idea that either service can stand in for the other at any time? | 19:07 |
| rraja | tbarron: I think so. | 19:08 |
| tbarron | rraja: if they want an active-active share service then that is not currently supported in manila, it's not a backend-specific issue. | 19:08 |
| *** absubram has joined #openstack-manila | 19:08 | |
| tbarron | rraja: this is parallel to the state of cinder today | 19:09 |
| tpsilva | vponomaryov1: hmm I'll see if I can find anything else | 19:09 |
| tbarron | rraja: you can run share service active/passive with a solution like pacemaker though | 19:09 |
| rraja | tbarron: hmmm. interesting. he said that he uses a similar configuration in Cinder. that's why I wanted to double check with you | 19:09 |
| openstackgerrit | Merged openstack/puppet-manila: Enable release notes translation https://review.openstack.org/383344 | 19:10 |
| tbarron | rraja: but I don't think that poses a problem for the assumption that only one manila service governs (does eviction, etc.) at any given time. | 19:10 |
| rraja | tbarron: http://pastebin.com/tTHPrxvu | 19:11 |
| tbarron | rraja: there *are* deployments out there that run cinder active-active, at their own risk, but upstream (and downstream at a place you and I know) it is not considered safe | 19:11 |
| tbarron | rraja: looks like they play tricks in cinder to have multiple cinder volume servers answer on the message queue as the same "host" | 19:13 |
| tbarron | rraja: there is ongoing work to make that kind of active-active cinder volume service safe but it has not yet landed | 19:13 |
| tbarron | rraja: but rather than argume with this person about cinder what can be said is that manila does not yet support active-active share service, so the current native cephfs implementation was not designed to support it either | 19:15 |
| rraja | tbarron: cool. got it. thanks a lot! :) | 19:16 |
| tbarron | rraja: np, and it's good that you brought this up here. road to active-active HA is one of the BCN design session topics | 19:16 |
| *** rraja has quit IRC | 19:26 | |
| *** xyang_ has quit IRC | 19:33 | |
| *** xyang_ has joined #openstack-manila | 19:35 | |
| *** lpetrut has quit IRC | 19:40 | |
| *** vponomaryov1 has quit IRC | 19:45 | |
| *** porrua has quit IRC | 19:48 | |
| openstackgerrit | Merged openstack/manila: TrivialFix: Remove default=None when set defaul value in Config https://review.openstack.org/377208 | 20:11 |
| *** merooney has quit IRC | 20:14 | |
| *** harlowja has joined #openstack-manila | 20:17 | |
| *** akerr has quit IRC | 20:22 | |
| openstackgerrit | Merged openstack/manila: Files with no code must be left completely empty https://review.openstack.org/378493 | 20:32 |
| *** dustins has quit IRC | 20:35 | |
| *** cknight has quit IRC | 20:52 | |
| *** timcl has quit IRC | 20:55 | |
| *** FL1SK has quit IRC | 20:55 | |
| *** ganso has quit IRC | 20:56 | |
| *** eharney has quit IRC | 21:09 | |
| *** harlowja has quit IRC | 21:17 | |
| *** timcl has joined #openstack-manila | 21:25 | |
| *** timcl has quit IRC | 21:26 | |
| *** gouthamr has quit IRC | 21:27 | |
| *** gouthamr has joined #openstack-manila | 21:31 | |
| *** gouthamr has quit IRC | 21:32 | |
| *** timcl has joined #openstack-manila | 21:35 | |
| *** gouthamr has joined #openstack-manila | 21:35 | |
| *** gouthamr has quit IRC | 21:37 | |
| *** harlowja has joined #openstack-manila | 21:39 | |
| *** harlowja has quit IRC | 21:39 | |
| *** harlowja has joined #openstack-manila | 21:39 | |
| *** merooney has joined #openstack-manila | 21:42 | |
| *** timcl has quit IRC | 21:49 | |
| *** absubram has quit IRC | 22:01 | |
| *** tpsilva has quit IRC | 22:04 | |
| *** assassin has quit IRC | 22:07 | |
| *** akshai has quit IRC | 22:07 | |
| *** ubijtsa has joined #openstack-manila | 22:08 | |
| *** ubijtsa is now known as Guest21331 | 22:08 | |
| *** dustins has joined #openstack-manila | 22:10 | |
| *** cknight has joined #openstack-manila | 22:10 | |
| *** xyang_ has quit IRC | 22:12 | |
| *** ociuhandu has quit IRC | 22:17 | |
| *** FL1SK has joined #openstack-manila | 22:21 | |
| *** merooney has quit IRC | 22:27 | |
| *** diablo_rojo has quit IRC | 22:33 | |
| *** gouthamr has joined #openstack-manila | 23:01 | |
| *** dustins has quit IRC | 23:10 | |
| *** cknight has quit IRC | 23:13 | |
| *** xyang1 has quit IRC | 23:18 | |
| *** faiz89 has joined #openstack-manila | 23:27 | |
| *** faiz89 has quit IRC | 23:27 | |
| *** faiz89 has joined #openstack-manila | 23:27 | |
| *** akapil has joined #openstack-manila | 23:33 | |
| *** akapil has quit IRC | 23:38 | |
| *** pcaruana has quit IRC | 23:55 | |
| *** csaba has quit IRC | 23:56 | |
| *** porrua has joined #openstack-manila | 23:58 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!