| *** LinPeiWen has joined #openstack-manila | 00:34 | |
| *** lkuchlan has joined #openstack-manila | 03:30 | |
| *** lkuchlan has quit IRC | 05:40 | |
| *** lkuchlan has joined #openstack-manila | 06:25 | |
| *** josecastroleon has joined #openstack-manila | 06:26 | |
| *** vishalmanchanda has joined #openstack-manila | 06:33 | |
| opendevreview | YuehuiLei proposed openstack/puppet-manila master: Changed minversion in tox to 3.18.0 https://review.opendev.org/c/openstack/puppet-manila/+/795032 | 08:04 |
|---|---|---|
| *** carloss has joined #openstack-manila | 10:43 | |
| *** nahimsouza has joined #openstack-manila | 11:14 | |
| *** enriquetaso has joined #openstack-manila | 11:14 | |
| zigo | Hi there! I've setup manila in my PoC cluster, and now, when starting manila-share, I have this in the logs: | 11:28 |
| zigo | manila.exception.BridgeDoesNotExist: Bridge br-int does not exist. | 11:28 |
| zigo | I don't understand why manila-share wants to lookup for br-int in my controller machines (which do *not* run OVS...). | 11:29 |
| zigo | Is there a way I can tell manila-share not to do that? | 11:29 |
| zigo | vkmc: ^ | 11:29 |
| *** toabctl has joined #openstack-manila | 11:33 | |
| *** tosky has joined #openstack-manila | 11:59 | |
| tbarron | zigo: are you running manila-share on a controller node and networking services on a separate dedicated node? | 12:06 |
| tbarron | zigo: the generic driver won't work in that deployment topology: https://bugs.launchpad.net/manila/+bug/1867950/comments/3 | 12:09 |
| opendevmeet | Launchpad bug 1867950 in OpenStack Shared File Systems Service (Manila) "[doc][rfe] Need to call out Generic Driver limitations" [Undecided,New] | 12:09 |
| zigo | tbarron: Ok, so it's probably advised to run manila-share on either a compute or a network node then? | 12:09 |
| tbarron | zigo: well *I* don't advise deploying it in production though we use it in gate for an open source software deployed back end that has features otherwise available only with proprietary backednds | 12:11 |
| tbarron | zigo: there the topology is all in one | 12:11 |
| tbarron | zigo: in production I run manila-share on controller nodes since currently it runs active-standby and we want exactly one instance running at all times | 12:13 |
| tbarron | and we use pacemaker for that | 12:13 |
| tbarron | zigo: just saying what I have experience with | 12:13 |
| zigo | tbarron: Does this mean I must run only a single instance of manila-share?!? | 12:13 |
| zigo | How can I make it HA ? | 12:14 |
| tbarron | zigo: but generic back end itself has no HA | 12:14 |
| zigo | Oh... :/ | 12:14 |
| zigo | So, I'd have to do that myself, I guess then. | 12:14 |
| *** fabiooliveira has joined #openstack-manila | 12:14 | |
| zigo | (I mean, the HA feature) | 12:14 |
| zigo | Thanks for all of the info. | 12:15 |
| tbarron | zigo: Yes that would be a way cool contribution :D | 12:15 |
| tbarron | zigo: it's not trivial though | 12:15 |
| zigo | tbarron: I was meaning, with corosync/pacemaker type of stuff... :) | 12:15 |
| tbarron | https://review.opendev.org/c/openstack/manila-specs/+/504987 | 12:16 |
| tbarron | zigo: ^^ | 12:16 |
| tbarron | zigo: getting a viable open source back end for manila is a continuing struggle. The only people *paid* to work on that are concentrating on ceph back end, at least for now, rather than generic. | 12:17 |
| zigo | tbarron: Is the CEPH backend fully HA, and multi-tenant ? | 12:18 |
| zigo | I'm fine with Ceph... | 12:18 |
| zigo | That would work for me. | 12:18 |
| tbarron | zigo: it is HA, bug if you run nfs in front of it currently that's HA only with pacemaker-corosync, like manila-share itself. | 12:19 |
| tbarron | Goal though is to get both of those running active-active without any pacemaker corosync required. | 12:19 |
| tbarron | On multi-tenancy, yes from a storage POV. If you and I are in separate keystone projects there is no way I can get your storage. | 12:20 |
| zigo | tbarron: Is there also the goal to use Cinder multi-attachment, so the share VM could be a double-VM with a shared VRRP port or something, like Octavia does? | 12:20 |
| tbarron | from networking POV it's more "subtle" -- there is one share server on a common share network (today) and that | 12:20 |
| tbarron | should have ebtables for no arp spoofing and client VMs should use a neutron policy that prevents ingress from anyone else on that network. | 12:21 |
| tbarron | In production our VMs use two nics, the regular one with private net and NAT, etc, and a second nic onto that storage network, which is itself unrouted. | 12:22 |
| tbarron | zigo: it's not a "goal" b/c no one has been developing the generic driver for a number of years now. Could be done probably. | 12:23 |
| zigo | tbarron: So, for production, you'd recommend using the Ceph driver then? | 12:24 |
| zigo | I mean, CephFS ... | 12:24 |
| zigo | Then no need to bother with the VMs and all, right? | 12:24 |
| tbarron | zigo: Currently I think that's the only viable open source production back end for Manila but | 12:27 |
| tbarron | other Manila folks might have other views or know something I don't. | 12:27 |
| tbarron | zigo: actually glusterfs is also pretty viable but the problem is that even though it is used in production the folks who use it | 12:28 |
| tbarron | aren't here for the most part and don't keep the upstream back end up to date | 12:28 |
| tbarron | Lately there have been a few new reviews for glusterfs though so maybe that situation will get better. | 12:29 |
| tbarron | zigo: But yes, if you use CephFS then you'll find others in #openstack-manila who work on it and keep it working. | 12:30 |
| tbarron | and who have customers who will yell at them if it breaks | 12:30 |
| tbarron | zigo: we also have a zfs on linux back end, but it too would need some love for production deployment. | 12:31 |
| tbarron | gotta run off now, will check in later .... | 12:32 |
| zigo | tbarron: Thanks a lot for all of the above. I'll try to setup CephFS then, but it lacks documentation. The doc only tells about the Generic backend ... :( | 12:48 |
| *** eharney has joined #openstack-manila | 13:06 | |
| zigo | manila-share doesn't need access to the db, right? | 13:25 |
| *** felipe_rodrigues has joined #openstack-manila | 13:56 | |
| *** caiquemello has joined #openstack-manila | 13:56 | |
| zigo | Looks like it needs it ... | 14:03 |
| opendevreview | Simon Dodsley proposed openstack/puppet-manila master: Support for Pure Storage flashBlade Manila driver https://review.opendev.org/c/openstack/puppet-manila/+/792848 | 14:04 |
| opendevreview | Simon Dodsley proposed openstack/puppet-manila master: Support for Pure Storage flashBlade Manila driver https://review.opendev.org/c/openstack/puppet-manila/+/792848 | 14:12 |
| *** lkuchlan has quit IRC | 14:21 | |
| opendevreview | Simon Dodsley proposed openstack/puppet-manila master: Support for Pure Storage flashBlade Manila driver https://review.opendev.org/c/openstack/puppet-manila/+/792848 | 14:26 |
| *** kafilat has joined #openstack-manila | 14:28 | |
| vkmc | hi all | 14:30 |
| vkmc | just a reminder that today we will start with our bug squash | 14:30 |
| vkmc | we will gather on https://meetpad.opendev.org/ManilaX-ReleaseBugSquash in 30min | 14:30 |
| vkmc | you can check more details in [Manila] Upcoming Bug Squash starting June 7th through June 11th 2021 (openstack-discuss mailing list) | 14:31 |
| vkmc | see you there! | 14:31 |
| opendevreview | Simon Dodsley proposed openstack/puppet-manila master: Support for Pure Storage flashBlade Manila driver https://review.opendev.org/c/openstack/puppet-manila/+/792848 | 14:49 |
| zigo | vkmc: Hi there ! Do you agree with tbarron's view that the generic driver shouldn't be used? | 14:49 |
| zigo | I'm getting "manila.exception.NoValidHost: No valid host was found. Failed to find a weighted host, the last executed filter was CapabilitiesFilter." in the scheduler's logs when creating a share btw, what should I look into? | 14:50 |
| vkmc | zigo, hi there! | 14:51 |
| tbarron | zigo: "shouldn't be used" is a bit strong, I just don't think it's suitable for *production* deployments if that means running with HA, etc. | 14:54 |
| tbarron | zigo: w.r.t DB, the share service does access the database | 14:54 |
| vkmc | zigo, well, there are many reasons for which we advise to look for other backend alternatives for production usage (https://docs.openstack.org/manila/latest/admin/generic_driver.html#known-restrictions) plus the ones that need to be | 14:55 |
| vkmc | documented | 14:55 |
| tbarron | zigo: and we have some cephfs back end doc here: https://docs.openstack.org/manila/latest/admin/cephfs_driver.html | 14:55 |
| vkmc | that were listed here https://bugs.launchpad.net/manila/+bug/1867950/ | 14:56 |
| opendevmeet | Launchpad bug 1867950 in OpenStack Shared File Systems Service (Manila) "[doc][rfe] Need to call out Generic Driver limitations" [Undecided,New] | 14:56 |
| opendevreview | Simon Dodsley proposed openstack/puppet-manila master: Support for Pure Storage flashBlade Manila driver https://review.opendev.org/c/openstack/puppet-manila/+/792848 | 14:56 |
| vkmc | zigo, for that error check the manila scheduler log to get more details, probably a capability missing from the share-type being used | 14:57 |
| tbarron | zigo: w.r.t. 'No valid host found' and scheduler capabilities filter you likely have mismatch between the capabilities (as extra specs) in the share type used to create the share and the capabilities that manila-share is reporting from the driver to the scheduler. | 14:57 |
| tbarron | zigo: debug logs for the manila scheduler should take you to the mismatch | 14:57 |
| tbarron | that is, what vkmc just said | 14:58 |
| vkmc | tbarron++ | 14:58 |
| vkmc | :) | 14:58 |
| tbarron | vkmc: as you know I am actually out today and will have to catch up with bug squash later | 14:58 |
| vkmc | tbarron, I am aware | 14:58 |
| vkmc | we will assign all bugs to you | 14:59 |
| tbarron | vkmc: i've got some bugs I'm working on and will add them to that list | 14:59 |
| vkmc | :D | 14:59 |
| tbarron | vkmc: of course :D | 14:59 |
| vkmc | cool! | 14:59 |
| zigo | vkmc: I get this: http://paste.openstack.org/show/806424/ | 15:01 |
| zigo | Though I really have: | 15:01 |
| zigo | [generic] | 15:01 |
| zigo | driver_handles_share_servers=True | 15:01 |
| zigo | in my manila-share server ... | 15:01 |
| tbarron | with debug on that log should show the actual capabilities reported, periodically. | 15:09 |
| tbarron | then as administrator run 'manila type-list'/'manila extra-specs-list' to see the "template" for extra-specs (and therefore back end capabilities) in the share type you are using when attempting the share create | 15:11 |
| tbarron | you may just need to update your share type definition | 15:11 |
| tbarron | ok, now I'm really out of here for the day | 15:11 |
| *** lkuchlan has joined #openstack-manila | 15:38 | |
| *** lkuchlan has quit IRC | 15:41 | |
| *** lkuchlan has joined #openstack-manila | 15:44 | |
| *** lkuchlan has quit IRC | 15:45 | |
| *** kiwi_36 has joined #openstack-manila | 16:23 | |
| *** lkuchlan has joined #openstack-manila | 16:50 | |
| *** lkuchlan has quit IRC | 16:51 | |
| *** kiwi_36_ has joined #openstack-manila | 17:30 | |
| *** vishalmanchanda has quit IRC | 19:12 | |
| *** lkuchlan has joined #openstack-manila | 19:36 | |
| *** lkuchlan has quit IRC | 19:36 | |
| *** kafilat has quit IRC | 19:47 | |
| opendevreview | Archana Kumari proposed openstack/python-manilaclient master: Add --wait flag to the delete share group operation https://review.opendev.org/c/openstack/python-manilaclient/+/789018 | 20:00 |
| opendevreview | Simon Dodsley proposed openstack/manila master: Add Pure Storage FlashArray driver https://review.opendev.org/c/openstack/manila/+/795180 | 20:36 |
| opendevreview | Simon Dodsley proposed openstack/manila master: WIP: Add Pure Storage FlashArray driver https://review.opendev.org/c/openstack/manila/+/795180 | 20:37 |
| opendevreview | Simon Dodsley proposed openstack/manila master: WIP: Add Pure Storage FlashArray driver https://review.opendev.org/c/openstack/manila/+/795180 | 21:08 |
| *** kiwi_36 has quit IRC | 21:12 | |
| opendevreview | Simon Dodsley proposed openstack/manila master: xWIP: Add Pure Storage FlashArray driver https://review.opendev.org/c/openstack/manila/+/795180 | 21:26 |
| *** caiquemello has quit IRC | 21:59 | |
| *** nahimsouza has quit IRC | 22:03 | |
| *** tosky has quit IRC | 22:04 | |
| *** fabiooliveira has quit IRC | 22:44 | |
| *** rraja has joined #openstack-manila | 23:11 | |
| *** enriquetaso has quit IRC | 23:36 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!