| *** jmlowe has quit IRC | 00:14 | |
| *** jmlowe has joined #openstack-manila | 00:15 | |
| openstackgerrit | Ben Swartzlander proposed openstack/manila master: Add NFSLocalHelper https://review.openstack.org/476239 | 00:23 |
|---|---|---|
| bswartz | zhongjun: https://review.openstack.org/#/c/444479/ | 00:43 |
| *** tuanluong has joined #openstack-manila | 00:50 | |
| *** kaisers__ has joined #openstack-manila | 01:28 | |
| *** kaisers_ has quit IRC | 01:31 | |
| *** chlong has joined #openstack-manila | 01:44 | |
| *** gregsfortytwo has quit IRC | 01:47 | |
| openstackgerrit | zhongjun proposed openstack/manila master: Use parenthesis instead of backslashes in tempest folder https://review.openstack.org/476027 | 01:50 |
| openstackgerrit | qtlu proposed openstack/manila master: Add support for Guru Meditation Reports for manila https://review.openstack.org/444688 | 02:00 |
| *** sapcc-bot1 has joined #openstack-manila | 02:00 | |
| *** david_ has quit IRC | 02:00 | |
| *** sapcc-bot has quit IRC | 02:00 | |
| openstackgerrit | qtlu proposed openstack/manila master: Add support for Guru Meditation Reports for manila https://review.openstack.org/444688 | 02:03 |
| *** chlong has quit IRC | 02:04 | |
| openstackgerrit | Vu Cong Tuan proposed openstack/manila master: Replace the usage of 'manager' with 'os_primary' https://review.openstack.org/476015 | 02:27 |
| *** chlong has joined #openstack-manila | 02:36 | |
| openstackgerrit | zhongjun proposed openstack/manila master: Use parenthesis instead of backslashes in tempest folder https://review.openstack.org/476027 | 02:49 |
| zhongjun | bswartz: Thanks a lot, I added some comments in this patch, please take a look | 03:06 |
| *** gouthamr has quit IRC | 03:39 | |
| zhongjun | gouthamr: Hi, It looks like we are only doing versioning in the shell or in the client when we add a new input field in this patch(https://review.openstack.org/#/c/468277/9/manilaclient/v2/shell.py) . Sometimes we're doing versioning in both the shell and in the client, because we need to display different things, and it could not work in the client. such as : | 03:45 |
| zhongjun | https://github.com/openstack/python-manilaclient/blob/master/manilaclient/v2/shell.py#L1169 | 03:45 |
| *** kaisers__ has quit IRC | 03:52 | |
| *** chlong has quit IRC | 04:02 | |
| *** chlong has joined #openstack-manila | 04:15 | |
| *** arnewiebalck_ has joined #openstack-manila | 04:16 | |
| *** kaisers_ has joined #openstack-manila | 04:52 | |
| *** kaisers_ has quit IRC | 05:24 | |
| *** jprovazn has joined #openstack-manila | 05:28 | |
| *** dsariel has quit IRC | 05:49 | |
| *** lpetrut has joined #openstack-manila | 06:03 | |
| *** kaisers_ has joined #openstack-manila | 06:20 | |
| *** dsariel has joined #openstack-manila | 06:36 | |
| *** kaisers_ has quit IRC | 06:36 | |
| *** pcaruana has joined #openstack-manila | 07:05 | |
| *** lpetrut_ has joined #openstack-manila | 07:11 | |
| *** lpetrut has quit IRC | 07:13 | |
| openstackgerrit | zhongjun proposed openstack/manila master: Add export-location filter in share and share instance list API https://review.openstack.org/461712 | 07:19 |
| *** kaisers_ has joined #openstack-manila | 07:34 | |
| *** a-pugachev has joined #openstack-manila | 07:44 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/manila-ui master: Imported Translations from Zanata https://review.openstack.org/476397 | 08:02 |
| *** sapcc-bot1 has quit IRC | 08:03 | |
| *** sapcc-bot has joined #openstack-manila | 08:03 | |
| *** david_ has joined #openstack-manila | 08:03 | |
| *** mkoderer_ has joined #openstack-manila | 08:03 | |
| *** seife_ has joined #openstack-manila | 08:03 | |
| *** tpatzig_ has joined #openstack-manila | 08:03 | |
| *** carthaca_ has joined #openstack-manila | 08:03 | |
| *** dgonzalez_ has joined #openstack-manila | 08:03 | |
| *** dgonzalez_ has quit IRC | 08:05 | |
| *** carthaca_ has quit IRC | 08:05 | |
| *** seife_ has quit IRC | 08:05 | |
| *** tpatzig_ has quit IRC | 08:05 | |
| *** mkoderer_ has quit IRC | 08:05 | |
| *** david_ has quit IRC | 08:05 | |
| *** kaisers_ has quit IRC | 08:07 | |
| *** sapcc-bot has quit IRC | 08:15 | |
| *** sapcc-bot2 has joined #openstack-manila | 08:15 | |
| *** dgonzalez_ has joined #openstack-manila | 08:16 | |
| *** david_ has joined #openstack-manila | 08:16 | |
| *** carthaca_ has joined #openstack-manila | 08:16 | |
| *** tpatzig_ has joined #openstack-manila | 08:16 | |
| *** seife_ has joined #openstack-manila | 08:16 | |
| *** mkoderer_ has joined #openstack-manila | 08:16 | |
| *** carthaca_ has quit IRC | 08:17 | |
| *** seife_ has quit IRC | 08:17 | |
| *** tpatzig_ has quit IRC | 08:17 | |
| *** dgonzalez_ has quit IRC | 08:17 | |
| *** david_ has quit IRC | 08:17 | |
| *** mkoderer_ has quit IRC | 08:18 | |
| openstackgerrit | Merged openstack/manila master: Retry backend initialization https://review.openstack.org/464205 | 08:29 |
| *** tpatzig_ has joined #openstack-manila | 08:30 | |
| *** dgonzalez_ has joined #openstack-manila | 08:30 | |
| *** carthaca_ has joined #openstack-manila | 08:30 | |
| *** mkoderer_ has joined #openstack-manila | 08:30 | |
| *** seife_ has joined #openstack-manila | 08:30 | |
| *** david_ has joined #openstack-manila | 08:30 | |
| *** tpatzig_ has quit IRC | 08:32 | |
| *** dgonzalez_ has quit IRC | 08:32 | |
| *** carthaca_ has quit IRC | 08:32 | |
| *** mkoderer_ has quit IRC | 08:32 | |
| *** seife_ has quit IRC | 08:32 | |
| *** david_ has quit IRC | 08:32 | |
| *** dmk0202 has joined #openstack-manila | 08:40 | |
| *** kaisers_ has joined #openstack-manila | 09:04 | |
| *** kaisers_ has quit IRC | 09:22 | |
| *** Administrator__ has joined #openstack-manila | 09:37 | |
| *** Administrator_ has quit IRC | 09:40 | |
| *** dmk0202 has quit IRC | 09:40 | |
| openstackgerrit | zhongjun proposed openstack/manila master: Add export-location filter in share and share instance list API https://review.openstack.org/461712 | 10:03 |
| *** kaisers_ has joined #openstack-manila | 10:19 | |
| zhongjun | vponomaryov: ping | 10:55 |
| *** ganso has joined #openstack-manila | 11:02 | |
| openstackgerrit | zhongjun proposed openstack/manila master: Add export-location filter in share and share instance list API https://review.openstack.org/461712 | 11:20 |
| openstackgerrit | Helen Walsh proposed openstack/manila master: VMAX VNX Manila - Refactor VMAX and VNX to use common code https://review.openstack.org/463106 | 11:22 |
| *** kaisers_ has quit IRC | 11:22 | |
| openstackgerrit | zhongjun proposed openstack/manila master: Enable IPv6 in manila(network plugins and drivers) https://review.openstack.org/406776 | 11:24 |
| openstackgerrit | Helen Walsh proposed openstack/manila master: VMAX VNX Manila - Refactor VMAX and VNX to use common code https://review.openstack.org/463106 | 11:25 |
| *** arnewiebalck_ has quit IRC | 11:27 | |
| openstackgerrit | Merged openstack/manila master: Replace the usage of 'manager' with 'os_primary' https://review.openstack.org/476015 | 11:35 |
| *** arnewiebalck_ has joined #openstack-manila | 11:39 | |
| vponomaryov | bswartz, tbarron: hello, you may be interested in following feature -> https://blueprints.launchpad.net/manila/+spec/generic-driver-boot-vm-from-volume | 11:54 |
| vponomaryov | zhongjun: pong | 11:54 |
| tbarron | vponomaryov: that may be useful as part of service VM HA solution | 11:58 |
| tbarron | vponomaryov: besides your use case | 11:58 |
| *** dsariel has quit IRC | 11:58 | |
| vponomaryov | tbarron: use case I am aware about is to be able to migrate VM while node is going down for maintenance | 11:58 |
| tbarron | vponomaryov: if the hypervisor for a service VM dies then it would be good to "nova evactuate" the SVM to another hypervisor | 11:59 |
| tbarron | vponomaryov: that requires shared backing store for the vm, or persistent shared volume | 11:59 |
| tbarron | that the vm boots from | 11:59 |
| vponomaryov | tbarron: so, you are interested in it, right? | 12:00 |
| tbarron | vponomaryov: persistent backing store for vm image via ceph in /var/lib/libvirt/images is good for some but cinder volume would be good for those that don't use ceph that way | 12:00 |
| tbarron | vponomaryov: yes | 12:00 |
| vponomaryov | tbarron: good to know, ty | 12:00 |
| *** tuanluong has quit IRC | 12:06 | |
| zhongjun | vponomaryov: Why we just remove some invalid filters for non-admin https://github.com/openstack/manila/blob/master/manila/api/common.py#L309 | 12:06 |
| *** rraja has joined #openstack-manila | 12:08 | |
| vponomaryov | zhongjun: I don't have strict opinion here. We inherited it from Cinder | 12:08 |
| zhongjun | vponomaryov: oh, I think we could do the same thing for admin and non-admin in our code | 12:11 |
| zhongjun | vponomaryov: https://review.openstack.org/#/c/461712/29/manila/api/v2/shares.py@434 Thanks for your review, As we did it before, we just pop the parameter in old API version when we add a new parameter in API. such as: we add 'is_public' in manage API. Do we really need to raise BadRequest in old API? | 12:15 |
| zhongjun | vponomaryov: such as: https://github.com/openstack/manila/blob/master/manila/api/v2/shares.py#L399 | 12:16 |
| zhongjun | vponomaryov: Do you have strict opinion here? | 12:17 |
| *** kaisers_ has joined #openstack-manila | 12:18 | |
| zhongjun | vponomaryov: such as: https://github.com/openstack/manila/blob/master/manila/api/v2/shares.py#L187 | 12:20 |
| vponomaryov | zhongjun: about this, my opinion is at https://review.openstack.org/#/c/461712/29/manila/api/v2/shares.py@434 removing arg and returning some data you make user think that his parameter was applied, but it is not indeed | 12:20 |
| *** gcb has joined #openstack-manila | 12:20 | |
| vponomaryov | zhongjun: and I consider it a bug | 12:21 |
| vponomaryov | zhongjun: no matter how many times it was done the same way before | 12:21 |
| zhongjun | vponomaryov: If it is a bug, it looks like we need to fix it in many places | 12:22 |
| openstackgerrit | Maurice Schreiber proposed openstack/manila master: Add ou to security service https://review.openstack.org/472154 | 12:24 |
| *** chlong has quit IRC | 12:26 | |
| *** rpittau__ is now known as rpittau | 12:31 | |
| *** eharney has quit IRC | 12:33 | |
| *** kaisers_ has quit IRC | 12:39 | |
| openstackgerrit | zhongjun proposed openstack/manila master: Enable IPv6 in manila(network plugins and drivers) https://review.openstack.org/406776 | 12:39 |
| *** cknight has joined #openstack-manila | 12:43 | |
| *** dsariel has joined #openstack-manila | 12:44 | |
| *** dustins has joined #openstack-manila | 13:15 | |
| *** jmlowe has quit IRC | 13:16 | |
| *** eharney has joined #openstack-manila | 13:18 | |
| bswartz | vponomaryov: that does look like an interesting feature | 13:19 |
| bswartz | vponomaryov: do we really need a stateful root volume though? the alternative is to keep all state in the Manila DB like we do today | 13:19 |
| *** catintheroof has joined #openstack-manila | 13:20 | |
| *** xyang1 has joined #openstack-manila | 13:26 | |
| *** gouthamr has joined #openstack-manila | 13:34 | |
| *** dsariel has quit IRC | 13:36 | |
| *** kaisers_ has joined #openstack-manila | 13:37 | |
| *** chlong has joined #openstack-manila | 13:39 | |
| openstackgerrit | Helen Walsh proposed openstack/manila master: VMAX VNX Manila - Refactor VMAX and VNX to use common code https://review.openstack.org/463106 | 13:40 |
| tbarron | bswartz: I was thinking a stateful root volume would be handy for migrating (or nova "evacuating") service VMs | 13:44 |
| tbarron | bswartz: stateful volume will at least in theory work for this even if the cloud deployer doesn't back /var/lib/libvirt/images with shared storage like nfs or rbd images | 13:45 |
| *** gouthamr has quit IRC | 13:47 | |
| openstackgerrit | Merged openstack/manila-ui master: Imported Translations from Zanata https://review.openstack.org/476397 | 13:52 |
| *** gouthamr has joined #openstack-manila | 13:52 | |
| bswartz | tbarron: doesn't nova migration take care of state in "ephemeral" root volumes for you? | 14:04 |
| tbarron | bswartz: I don't think so, I think you have to set up shared storage for the VM images or boot from cinder volume | 14:05 |
| bswartz | tbarron: that's interesting | 14:07 |
| bswartz | hooray for boot-from-volume | 14:07 |
| bswartz | tbarron: does nova actually understand the concept of shared "ephemeral" storage in some explicit way? | 14:08 |
| * smcginnis laughs | 14:08 | |
| bswartz | smcginnis: ^ do you know the answer to this? | 14:09 |
| smcginnis | :) | 14:09 |
| smcginnis | bswartz: I believe the answer is no. | 14:09 |
| tbarron | bswartz: my impression is that it just migrates the VM (or "evacuates" it) and that if there is shared backing store | 14:09 |
| bswartz | if not, then how could nova take advantage of it during a migration? | 14:09 |
| *** lpetrut_ has quit IRC | 14:09 | |
| tbarron | when the VM comes up on the other side the kept state is available | 14:10 |
| bswartz | does it just rely on file name uniqueness, and look if the file is there, and if so, assumes that it's shared? | 14:10 |
| tbarron | in the case of an "evacuate" at least | 14:10 |
| tbarron | it has to start the VM up on the new hypervisor, right? | 14:11 |
| bswartz | AIUI, there's a period of time where the VM exists in both places | 14:11 |
| tbarron | bswartz: I'm thinking more of the "evacuate" case, where the source hypervisor has failed | 14:12 |
| bswartz | it's not like star trek teleportation -- it's more like when you "mv" a file across filesystems -- it's really a "cp" followed by a "rm" | 14:12 |
| bswartz | oh | 14:13 |
| tbarron | bswartz: "instance" HA case | 14:13 |
| bswartz | you're talking about resurrection of dead VMs? | 14:13 |
| tbarron | bswartz: yes, resurrect would have been a better word than "evacuate" | 14:14 |
| tbarron | nova "evacuates" the victims after they've already been killed once | 14:14 |
| bswartz | or reincarnation may be the more apt analogy, since it will live in a new body | 14:14 |
| tbarron | in traditional christian and jewish theologies resurrected people had some kind of body | 14:14 |
| tbarron | post-Descartes that idea tended to go away | 14:15 |
| tbarron | but that's an aside | 14:15 |
| bswartz | evacuate is a term I would use _before_ the hypervisor has died -- when you plan to bring it down for maintenance and you're relocating the still-living VMs | 14:15 |
| tbarron | bswartz: I agree, but the command is called "nova evacuate" - go figure | 14:15 |
| bswartz | I'm only familiar with the Hindu/Buddhist notions of resurrection | 14:15 |
| bswartz | s/resurrection/reincarnation/ | 14:16 |
| tbarron | where the goal is to quit getting resurrected because it's the same damn thing over and over again | 14:16 |
| tbarron | s/resurrected/reincarnated/ | 14:16 |
| tbarron | kinda like OpenStack discusssions | 14:17 |
| bswartz | ROFL | 14:17 |
| bswartz | okay so back to something relevant -- how does this notion of evacuation/VM reincarnation apply to manila? | 14:17 |
| tbarron | bswartz: I'm interested in eliminating SPOF in data path when using service VMs | 14:18 |
| bswartz | so are you concerned about the ability to move a living service VM or what to do about dead service VMs? | 14:18 |
| tbarron | so resurrecting a VM on another hypervisor when the original node fails is part of the solution | 14:18 |
| tbarron | bswartz: more the second | 14:18 |
| tbarron | the problem is hard | 14:18 |
| tbarron | perhaps the simplest place to start is with a physical failure of a node hosting a service VM | 14:19 |
| bswartz | I disagree -- although I admin not having thought deeply about this topic | 14:19 |
| bswartz | admit | 14:19 |
| bswartz | My intuition is that all of the state to reincarnate the service VM must live in the manila DB | 14:19 |
| bswartz | therefore there's no value in preserving the root volume of the service image | 14:20 |
| tbarron | we need to detect that automatically, fence the host (in case it's not really gone but isn't reachable); and "evacuate" to a new host | 14:20 |
| tbarron | and we need a way to set that up with installer tools | 14:20 |
| tbarron | or have manila driver set it up | 14:20 |
| bswartz | oh god you have to consider split brain problems here | 14:21 |
| tbarron | bswartz: you raise another interesting point though; do we have to rely on nova for the re-incarnation-with-state | 14:21 |
| tbarron | in some ways the last part (boot up on new hypervisor with state) may be the easiest part | 14:21 |
| tbarron | bswartz: but your point needs to be understood anyways, if with a manila SVM DB state is sufficient and one doesn't need | 14:22 |
| *** dsariel has joined #openstack-manila | 14:22 | |
| bswartz | I has assumed the HA scheme was going to look more like an active-active scheme where you had 2 running service images that were in contact with eachother, such that either one could die without loss of data or availability to the client | 14:22 |
| tbarron | shared state for the SVM itself outside the DB, that's worth understanding | 14:22 |
| bswartz | in the case I described, the more interesting question is what to do when your active node count drops from 2 to 1 | 14:23 |
| tbarron | bswartz: I am playing with that idea too, but that requires a VIP that migrates | 14:23 |
| bswartz | when your active node count drops from 1 to 0, then the obvious thing to do is to build a new one from scratch, using the Manila DB state to configure the new service image | 14:24 |
| tbarron | bswartz: in my case I want ganesha in the service VMs and ganesha won't run active-active yet | 14:25 |
| bswartz | if there's anything in that service image which isn't persisted in the manila DB, then I think we have a problem | 14:25 |
| bswartz | what do you mean by "yet"? | 14:25 |
| bswartz | is it a feature under active development? planned for the future? works today but with limitations? | 14:26 |
| tbarron | bswartz: but even if it did, you want to keep the export location from the dead service VM still working | 14:26 |
| tbarron | bswartz: I repeat a hand wave :) don't think it's being actively developed | 14:26 |
| bswartz | the point of multiple export locations combined with pNFS is that it doesn't matter if one of the export locations is down -- you can just use another | 14:27 |
| tbarron | bswartz: so using one SVM and evactuating/migrating/reincarnating it is attractive | 14:27 |
| tbarron | bswartz: which one did the user VM put in its fstab? | 14:28 |
| bswartz | well I agree if neither nfs-kernel-server or ganesha-nfs has proper pNFS support, then you're stuck with some kind of active/passive scheme | 14:28 |
| tbarron | bswartz: it *could* just use another, but how would it know to do so? | 14:28 |
| tbarron | bswartz: dumb question, does pNFS fix that issue from the user VM/client side? | 14:29 |
| bswartz | tbarron: it does after the mount succeeds | 14:29 |
| bswartz | you're correct that the node that happens to be down is the IP in your fstab then you have a problem | 14:30 |
| bswartz | I suspect this is why netapp does IP takeover for NFS | 14:30 |
| tbarron | bswartz: ^^ | 14:30 |
| tbarron | bswartz: yeah, that's what I figured re. NetApp too | 14:30 |
| bswartz | does nova have a facility for IP takeover? | 14:30 |
| bswartz | could you grab the neutron port and reassign it to a different nova VM? | 14:31 |
| tbarron | bswartz: evacuate moves the VM to new hypervisor and it comes up with the same IPs | 14:31 |
| bswartz | oh I see where you're going | 14:31 |
| tbarron | bswartz: I don't *know* how well this interacts with neutron though, am researching | 14:31 |
| tbarron | I think it's supposed to work and there may be some bugs | 14:31 |
| bswartz | this isn't even active/passive, this is just automatically restarting | 14:32 |
| tbarron | bswartz: right, I was mixing topics | 14:32 |
| bswartz | okay let me consider that scenario | 14:32 |
| tbarron | bswartz: I'm exploring several fronts. | 14:32 |
| tbarron | original subject was migration/evacuation/reincarnation of a single service VM | 14:33 |
| *** jmlowe has joined #openstack-manila | 14:33 | |
| tbarron | to a new node/hypervisor | 14:33 |
| bswartz | so the theory is that even when the hypervisor is gone, nova remembers the old VM and its neutron ports etc | 14:33 |
| tbarron | doesn't matter if it's ganesha or generic or whatever | 14:33 |
| bswartz | and it's easier to ask nova to just bring the same old instance up somewhere else than it is to create a new instance from scratch using the same image ID and neutron ports | 14:33 |
| tbarron | yeah, it's a theory/conjecture to explore | 14:34 |
| bswartz | and the main challenge in that case is what to do about the root volume, because it's most likely gone in the ephemeral case, but still available in the cinder volume case | 14:34 |
| tbarron | it's what people who are working on "instance HA" for "pet" VMs are looking at | 14:34 |
| bswartz | in the ephemeral case, you could in theory reconstruct the whole service VM after it booted up from the image with a clean state | 14:35 |
| tbarron | well, that's a challenge (others are detection, fencing, etc. but we'll wave hands on those for the moment since they are there whichever approach for re-gaining state is taken) | 14:35 |
| bswartz | the only state that would need restoring would be all of the shares and their access rules | 14:35 |
| tbarron | yeah, there's the manila state and the neutron/port state. | 14:37 |
| bswartz | the ephemeral case seems cleaner to me because you have one source of truth | 14:37 |
| tbarron | so you are arguing atm for using ensure-shares to restore the shares, access-rules, and exports ... | 14:38 |
| bswartz | tbarron: honestly I'm not sure what's required on the manila driver side | 14:41 |
| bswartz | but it strikes me that we should be able to solve this problem without resorting to boot from volume for service images | 14:42 |
| bswartz | and it feel intuitively like the more robust soluition | 14:42 |
| bswartz | in the end it might be more work if we have to add more driver interfaces | 14:43 |
| bswartz | I'd need to think about what the manila side of the solution would look like -- who is watching for heartbeats and how are we pulling the trigger on a reincarnation process | 14:44 |
| tbarron | bswartz: I do think there are multiple approaches for the manila state and wasn't looking at boot-from-volume (or shared backing store for libvirt images) from thinking about that issue per se | 14:44 |
| tbarron | bswartz: it was more looking at SVM HA in the context of current approaches in OpenStack for "instance HA" | 14:45 |
| tbarron | bswartz: and wondering if that could/should be leveraged | 14:45 |
| tbarron | bswartz: which other projects than manila have SVMs that are currently SPOFs for data path (or other critical service path)? | 14:46 |
| *** markstur has joined #openstack-manila | 14:52 | |
| bswartz | we should be clear that it's only certain drivers that suffer from this limitation | 14:59 |
| bswartz | manila is designed to be out of the data path | 14:59 |
| tbarron | bswartz: right, but open source drivers that use SVMs have this issue | 15:00 |
| bswartz | tbarron: the same could be true of any driver in any project that leveraged nova | 15:01 |
| bswartz | just saying that drivers are a special case | 15:01 |
| tbarron | any project that uses OpenStack VMs, or OpenStack containers to build an open source "appliance" that delivers the core goods | 15:02 |
| *** MVenesio has joined #openstack-manila | 15:08 | |
| *** sapcc-bot2 has quit IRC | 15:10 | |
| *** sapcc-bot has joined #openstack-manila | 15:10 | |
| *** eharney_ has joined #openstack-manila | 15:17 | |
| *** eharney has quit IRC | 15:18 | |
| *** kaisers_ has quit IRC | 15:19 | |
| *** gregsfortytwo has joined #openstack-manila | 15:22 | |
| *** cknight has quit IRC | 15:25 | |
| *** eharney_ has quit IRC | 15:30 | |
| *** rraja has quit IRC | 15:31 | |
| *** eharney has joined #openstack-manila | 15:34 | |
| *** jmlowe_ has joined #openstack-manila | 15:36 | |
| *** jmlowe has quit IRC | 15:38 | |
| *** jprovazn has quit IRC | 15:38 | |
| *** pcaruana has quit IRC | 15:42 | |
| openstackgerrit | Merged openstack/manila master: Use parenthesis instead of backslashes in tempest folder https://review.openstack.org/476027 | 15:48 |
| tbarron | ^^ such projects use VMs (or other kinds of compute) as stateful "pets" and face a challenge of what to do if the host for the pet dies | 15:55 |
| * tbarron ignores more subtle failures at the moment | 15:56 | |
| tbarron | bswartz: I'm going to get lunch but just followed up with another remark, I had stopped so as not to interrupt the communnity meeting | 15:57 |
| openstackgerrit | Ben Swartzlander proposed openstack/manila master: Add NFSLocalHelper https://review.openstack.org/476239 | 16:01 |
| bswartz | tbarron: I don't disagree -- just pointing out that manila is not unique in this regard | 16:02 |
| bswartz | I'm sensitive to suggestions that manila is in the data path | 16:02 |
| bswartz | drivers may or may not cause other openstack components to be in the data path | 16:03 |
| bswartz | but of course drivers can do practically *anything* | 16:03 |
| tbarron | bswartz: ack | 16:05 |
| tbarron | bswartz: ignoring the manila data service manila itself is not in the data path | 16:06 |
| tbarron | bswartz: but some of its drivers rely on SVMs and the like which are in the data path | 16:08 |
| tbarron | bswartz: and which may be a SPOF | 16:09 |
| tbarron | bswartz: as long as it is understood that those drivers are for reference/research purposes, cool | 16:10 |
| vponomaryov | tbarron: https://www.openstack.org/assets/survey/April2017SurveyReport.pdf page 59 | 16:15 |
| *** cknight has joined #openstack-manila | 16:15 | |
| *** kaisers_ has joined #openstack-manila | 16:16 | |
| vponomaryov | tbarron: so, generic driver definitely not "for reference/research purposes only" | 16:18 |
| tbarron | vponomaryov: yeah, one fifth are using it and it introduces a SPOF in the data plane :( | 16:22 |
| tbarron | vponomaryov: so it is a problem worth solving | 16:22 |
| *** cknight has left #openstack-manila | 16:22 | |
| tbarron | vponomaryov: cephfs driver doesn't have this issue but ganesha in front of cephfs will if we put it in a SVM | 16:22 |
| tbarron | and zfs on linux has it in that circumstance | 16:23 |
| tbarron | i'd like a robust software only open source solution | 16:23 |
| vponomaryov | tbarron: also, if you look at page 51, you will see, that the majority is openvswitch - the only L2 element supported and tested with generic driver | 16:25 |
| vponomaryov | plus multitenancy that generic driver has | 16:25 |
| vponomaryov | all that in summary means that this driver covers really big field of use cases | 16:26 |
| tbarron | vponomaryov: agree | 16:26 |
| tbarron | vponomaryov: so lots of use cases but not robust enough for production as long as it introduces a SPoF in data plane | 16:28 |
| tbarron | vponomaryov: cloud deployer might not care but supporting a distro I have to care | 16:29 |
| vponomaryov | tbarron: it is question to cinder's backend | 16:29 |
| vponomaryov | tbarron: losing Vm is only downtime, but not losing data | 16:30 |
| tbarron | vponomaryov: even if cinder backend is completely robust you still have SPoF b/c SVM is a single unprotected nova instance | 16:30 |
| tbarron | vponomaryov: oh, I agree data loss is more serious than downtime | 16:31 |
| tbarron | vponomaryov: but downtime on data path is also a sufficiently big issue | 16:31 |
| tbarron | vponomaryov: and we have open source solutions for the data loss/integrity issue | 16:32 |
| *** kaisers_ has quit IRC | 16:46 | |
| *** a-pugachev has quit IRC | 16:49 | |
| openstackgerrit | Ben Swartzlander proposed openstack/manila master: Add NFSLocalHelper https://review.openstack.org/476239 | 17:25 |
| bswartz | after I tackle LVM I plan to make the generic driver great again | 17:26 |
| *** kaisers_ has joined #openstack-manila | 17:42 | |
| *** arnewiebalck_ has quit IRC | 17:47 | |
| openstackgerrit | Victoria Martinez de la Cruz proposed openstack/manila master: Add create/delete/extend/shrink share notifications https://review.openstack.org/466586 | 17:59 |
| *** eharney has quit IRC | 18:25 | |
| *** kaisers_ has quit IRC | 18:25 | |
| *** sapcc-bot has quit IRC | 18:29 | |
| *** sapcc-bot has joined #openstack-manila | 18:29 | |
| *** jprovazn has joined #openstack-manila | 18:42 | |
| *** arnewiebalck_ has joined #openstack-manila | 18:55 | |
| *** kaisers_ has joined #openstack-manila | 18:56 | |
| *** eharney has joined #openstack-manila | 18:57 | |
| *** jprovazn has quit IRC | 19:38 | |
| *** bswartz has quit IRC | 19:38 | |
| *** MVenesio has quit IRC | 20:08 | |
| *** a-pugachev has joined #openstack-manila | 20:16 | |
| *** eharney has quit IRC | 20:18 | |
| *** kaisers_ has quit IRC | 20:39 | |
| *** kaisers_ has joined #openstack-manila | 20:39 | |
| *** jmlowe_ has quit IRC | 20:41 | |
| *** arnewiebalck_ has quit IRC | 20:58 | |
| *** arnewiebalck_ has joined #openstack-manila | 20:59 | |
| *** arnewiebalck_ has quit IRC | 21:00 | |
| *** dmellado has quit IRC | 21:00 | |
| *** dmellado has joined #openstack-manila | 21:03 | |
| *** dustins has quit IRC | 21:14 | |
| *** catintheroof has quit IRC | 21:18 | |
| *** a-pugachev has quit IRC | 21:19 | |
| *** kaisers_ has quit IRC | 21:36 | |
| *** kaisers_ has joined #openstack-manila | 21:41 | |
| *** kaisers_ has quit IRC | 21:47 | |
| *** gouthamr has quit IRC | 21:55 | |
| *** xyang1 has quit IRC | 21:56 | |
| *** chlong has quit IRC | 21:57 | |
| *** kaisers_ has joined #openstack-manila | 22:07 | |
| *** gouthamr has joined #openstack-manila | 22:16 | |
| *** ganso has quit IRC | 22:29 | |
| *** dmellado has quit IRC | 22:33 | |
| *** dmellado has joined #openstack-manila | 22:37 | |
| *** dmellado has quit IRC | 23:03 | |
| *** dmellado has joined #openstack-manila | 23:06 | |
| *** chlong has joined #openstack-manila | 23:10 | |
| *** dmellado has quit IRC | 23:15 | |
| *** dmellado has joined #openstack-manila | 23:18 | |
| *** bswartz has joined #openstack-manila | 23:25 | |
| *** dmellado has quit IRC | 23:31 | |
| *** winston-d_ has joined #openstack-manila | 23:44 | |
| *** dmellado has joined #openstack-manila | 23:51 | |
| *** dmellado has quit IRC | 23:55 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!