| *** amotoki has quit IRC | 00:02 | |
| *** salv-orlando has quit IRC | 00:25 | |
| *** yamamoto_ has joined #openstack-neutron-release | 01:44 | |
| *** yamahata has quit IRC | 01:56 | |
| *** iyamahat has quit IRC | 01:57 | |
| *** amotoki has joined #openstack-neutron-release | 01:58 | |
| *** amotoki has quit IRC | 02:02 | |
| *** rossella_s has quit IRC | 02:03 | |
| *** rossella_s has joined #openstack-neutron-release | 02:03 | |
| *** yamamoto_ has quit IRC | 02:04 | |
| * carl_baldwin here | 02:40 | |
| HenryG | armax: join us? | 02:41 |
|---|---|---|
| armax | HenryG: we have beer | 02:41 |
| HenryG | what?! where? | 02:41 |
| armax | right now the create_subnet API | 02:42 |
| armax | can figure out the default subnet pool if no cidr is specified | 02:42 |
| armax | what if we added another arg, like default=True | 02:43 |
| armax | like here: | 02:43 |
| armax | https://github.com/openstack/neutron/blob/master/neutron/services/auto_allocate/db.py#L211 | 02:43 |
| carl_baldwin | So, like: neutron subnet-create --default-subnet-pool ... ? | 02:44 |
| armax | carl_baldwin: yes | 02:44 |
| armax | that would imply a tweak to the extension | 02:44 |
| armax | not 100% convinced yet | 02:45 |
| carl_baldwin | One thing to consider is that the default subnet pool option was originally added so that operators could turn on prefix delegation for everyone automatically. They did this by specifying a special subnet pool for PD. | 02:45 |
| carl_baldwin | I liked that PD subnets showed like they were in a subnet pool. It will make them address scope compatible. | 02:46 |
| armax | ic so the route to explicitly ask for a default subnet-pool is not really viable | 02:47 |
| carl_baldwin | Not sure yet. | 02:47 |
| *** iyamahat has joined #openstack-neutron-release | 02:47 | |
| carl_baldwin | My idea to change the tempest tests to pass None for subnetpools may not be viable either until tempest tests no longer have to work for Juno. | 02:48 |
| *** yamahata has joined #openstack-neutron-release | 02:48 | |
| armax | however | 02:49 |
| armax | we hit the failure when a cidr is indeed specified | 02:49 |
| armax | wouldn’t that imply subnetpool=None? | 02:50 |
| armax | perhaps that’s our answer | 02:50 |
| armax | I mean | 02:50 |
| armax | if the user does specify a cidr and no subnetpool | 02:50 |
| carl_baldwin | John Davidge (I think) proposed that and I shot it down at the time. | 02:51 |
| armax | wouldn’t we assume subnetpool=None | 02:51 |
| carl_baldwin | I didn't like that the presence of the cidr in the API changed the subnetpool that was used (or not) | 02:51 |
| *** yamamoto_ has joined #openstack-neutron-release | 02:51 | |
| carl_baldwin | It seemed a really odd side-effect to me. | 02:51 |
| HenryG | Couldn't that allow one tenant to mess up routing for other tenants? | 02:52 |
| carl_baldwin | HenryG: How? | 02:52 |
| armax | carl_baldwin: it depends on how you look at it | 02:52 |
| *** salv-orlando has joined #openstack-neutron-release | 02:53 | |
| armax | I can still do subnet-create cidr —subnetpool=None and that would have the same effects of subnet-create cidr with the revised semantic | 02:53 |
| armax | but it is a bit odd | 02:53 |
| armax | we could have a validator that makes that more explicit for us | 02:53 |
| armax | maybe not | 02:55 |
| carl_baldwin | I expect that if I add --cidr <something> to any subnet create command that the subnet I get is from the same pool as if I didn't add it. | 02:55 |
| *** salv-orlando has quit IRC | 02:55 | |
| armax | assuming that | 02:56 |
| armax | I may have subnets already associated to a subnet pool | 02:56 |
| armax | but i see what you’re saying | 02:56 |
| armax | there’s a danger that the behavior of the API changes based on the order of the request | 02:57 |
| armax | s | 02:57 |
| carl_baldwin | The presence or lack thereof of --cidr on the command line to me say whether I want to try to choose something or accept what the system allocates. It would be a surprise if it also affected what pool the subnet is associated with. | 02:57 |
| carl_baldwin | And, since address scopes are based on subnet pools, it would also affect the address scope that the subnet is in. | 02:58 |
| armax | however | 02:58 |
| armax | today if I specify just the cidr and I don’t have default subnet pools | 02:58 |
| armax | what happens? | 02:58 |
| armax | (but I have more than one) | 02:58 |
| carl_baldwin | More than one what? | 02:59 |
| armax | pool | 02:59 |
| carl_baldwin | But none of the pools is default? | 02:59 |
| armax | yes | 02:59 |
| carl_baldwin | You get a subnet from the old pool. The legacy pool. Whatever you want to call it. The one that allows overlaps and arbitrary usage of addresses. | 03:00 |
| carl_baldwin | You get the old behavior. | 03:00 |
| armax | so the call is equivalent to passing subnetpool=None | 03:01 |
| armax | then | 03:01 |
| carl_baldwin | es. | 03:01 |
| carl_baldwin | Yes. | 03:01 |
| carl_baldwin | Basically, the code looks up the default subnet pool which returns None. So, they are exactly equivalent. | 03:01 |
| armax | I see | 03:02 |
| armax | I can’t think of anything right now | 03:03 |
| armax | the other not ideal solution would be t | 03:03 |
| carl_baldwin | It was that realization that led me to propose having existing tempest tests pass None by default. | 03:03 |
| armax | to have an explicit config for gman that tells me what pool I should pick from | 03:03 |
| armax | but I don’t even wanna go there | 03:04 |
| armax | true, but this isn’t just about tests | 03:04 |
| carl_baldwin | Do tempest tests need to support Juno or earlier now? | 03:04 |
| armax | I think juno was droppe | 03:04 |
| armax | d | 03:04 |
| carl_baldwin | I know it isn't just about tests. But, if we can figure out the right way to test it, that might lead us to the right solution. Or, vice-versa. | 03:05 |
| armax | dunno if it’s on some periodic queue though | 03:05 |
| armax | there’s some bit-rot set of jobs that run periodically I believe | 03:05 |
| carl_baldwin | I can chase that down if we think this option is worth considering. | 03:06 |
| armax | not sure it is | 03:06 |
| armax | not yet anyway | 03:07 |
| carl_baldwin | I think there is a fundamental problem that makes it hard. | 03:08 |
| carl_baldwin | We're trying to provide a path for people to transition to the wild-wild-west of addressing where tenants supply whatever they want to a more controlled form of addressing where operators provision pools. | 03:09 |
| carl_baldwin | ... while supporting the wild-wild-west way of doing things. | 03:09 |
| armax | yeah | 03:09 |
| carl_baldwin | I'm not saying that we shouldn't support both ways. I think we have to. | 03:09 |
| armax | the hard part is to figure out how the two co-exist | 03:09 |
| carl_baldwin | Both must be supported. | 03:09 |
| armax | I can’t imagine that one supplants the other | 03:10 |
| carl_baldwin | But, it makes for some awkward stuff like that. | 03:10 |
| carl_baldwin | s/that/this. | 03:10 |
| armax | because overlapping IP’s a is a great selling proposition to net-virt | 03:10 |
| carl_baldwin | I didn't quite understand that last sentence. | 03:10 |
| armax | overlapping IPs is synomous of true multi-tenancy and network virtualization | 03:11 |
| carl_baldwin | Well, subnetpools supports overlapping IPs in a much better way, IMO. Use more than one pool. | 03:12 |
| armax | ya | 03:13 |
| armax | the problem stems from the default one | 03:13 |
| carl_baldwin | Right. | 03:13 |
| armax | so perhaps | 03:14 |
| HenryG | That bans all wild west activities? | 03:14 |
| carl_baldwin | It was never my intention to prevent address overlap altogether. Just to control where it is allow to happen.d | 03:14 |
| armax | perhaps | 03:14 |
| carl_baldwin | HenryG: No, not bans. | 03:14 |
| carl_baldwin | HenryG: It just makes you explicitly request the wild-wild-west by passing --subnetpool=None | 03:14 |
| armax | I can’t create overlapping subnetpools by any chance? | 03:15 |
| carl_baldwin | I think my little brain might be brewing another option. | 03:15 |
| armax | :) | 03:15 |
| carl_baldwin | Define overlapping subnetpools | 03:15 |
| carl_baldwin | Two subnet pools with overlapping addresses? Or, a subnet pool that allows overlap within it? | 03:16 |
| armax | well | 03:17 |
| carl_baldwin | Here is the new option. We blow away the global default subnet pool and we add a default subnet pool to the external network. | 03:17 |
| armax | I could do this | 03:17 |
| carl_baldwin | My new option might need some more baking. | 03:17 |
| armax | not sure I follow | 03:18 |
| armax | but it looks like I can create overlapping subnetpools | 03:18 |
| armax | eg. | 03:18 |
| armax | I could do | 03:18 |
| armax | neutron subnetpool-create --default-prefixlen 24 --pool-prefix 10.0.0.0/16 test-pool | 03:18 |
| armax | a number of times | 03:18 |
| armax | within the context of the same tenant or across tenants | 03:18 |
| armax | carl_baldwin: do you concur? | 03:18 |
| carl_baldwin | Yes, that was always in the design. Like I said, I never wanted to prevent overlapping addresses. | 03:18 |
| armax | now | 03:18 |
| armax | I can also have a default subnetpool | 03:19 |
| * carl_baldwin not seeing how it helps yet. | 03:19 | |
| armax | hang on | 03:19 |
| armax | I may have just brainfarted | 03:19 |
| armax | how would the default subnetpool on the external network help? | 03:20 |
| carl_baldwin | I'm not sure that works either. | 03:21 |
| armax | I think we have to sleep on it | 03:21 |
| armax | if we had an answer, a good one, we’d have it by now | 03:22 |
| carl_baldwin | I agree. My favorite is still changing tempest to be explicit about testing the wild-wild-west. That's really what it is doing with all the tests that aren't explicitly testing subnet pools. | 03:23 |
| carl_baldwin | Let's talk again later. | 03:23 |
| carl_baldwin | Even though its my favorite, I'm not totally sold on it myself. | 03:24 |
| armax | that won’t tackle the non-backward compatible of creating subnets | 03:24 |
| carl_baldwin | Right (I think I know what you mean) | 03:25 |
| armax | and that really breaks with default subnetpools, does it not? | 03:25 |
| armax | I mean in presence of | 03:26 |
| carl_baldwin | It does. I was just hoping to provide a way for an operator to break free from the wild-wild-west without having to carry around an extra option to subnet-create everywhere. | 03:27 |
| armax | so I am just have gonna have you elaborate more why falling back on subnetpool=None in presence of a cidr | 03:27 |
| armax | but I think I am halfway there | 03:27 |
| carl_baldwin | How about next time? | 03:28 |
| armax | yup | 03:28 |
| carl_baldwin | It is 8:30 here and I've got to say goodnight to the kids. | 03:28 |
| armax | I am drained anyway | 03:28 |
| carl_baldwin | Me too. | 03:29 |
| carl_baldwin | Good discussions today. Thanks. | 03:29 |
| armax | and it’s even later for HenryG if he hasn’t fallen asleep on the keyboard | 03:29 |
| armax | carl_baldwin: tahnks! | 03:29 |
| armax | ttyl | 03:29 |
| HenryG | Thanks guys, I'll ping you tomorrow | 03:29 |
| armax | nite HenryG | 03:29 |
| *** yamahata has quit IRC | 04:17 | |
| *** salv-orlando has joined #openstack-neutron-release | 04:37 | |
| *** amotoki has joined #openstack-neutron-release | 04:41 | |
| *** salv-orlando has quit IRC | 04:42 | |
| armax | ……..0 | 04:47 |
| *** iyamahat has quit IRC | 05:01 | |
| *** iyamahat_ has joined #openstack-neutron-release | 05:01 | |
| *** salv-orlando has joined #openstack-neutron-release | 05:54 | |
| *** salv-orlando has quit IRC | 05:56 | |
| *** rossella_s has quit IRC | 06:03 | |
| *** rossella_s has joined #openstack-neutron-release | 06:03 | |
| *** iyamahat_ has quit IRC | 06:33 | |
| *** armax has quit IRC | 06:54 | |
| *** salv-orlando has joined #openstack-neutron-release | 07:12 | |
| *** amotoki has quit IRC | 07:16 | |
| *** salv-orlando has quit IRC | 07:23 | |
| *** amotoki has joined #openstack-neutron-release | 07:24 | |
| *** amotoki has quit IRC | 07:35 | |
| *** amotoki has joined #openstack-neutron-release | 07:39 | |
| *** amotoki has quit IRC | 07:43 | |
| *** amotoki has joined #openstack-neutron-release | 07:55 | |
| *** salv-orlando has joined #openstack-neutron-release | 08:28 | |
| *** salv-orlando has quit IRC | 08:35 | |
| *** ig0r_ has quit IRC | 08:45 | |
| *** salv-orlando has joined #openstack-neutron-release | 09:05 | |
| *** amotoki has quit IRC | 09:34 | |
| *** amotoki has joined #openstack-neutron-release | 09:48 | |
| *** salv-orlando has quit IRC | 09:52 | |
| *** amotoki has quit IRC | 09:53 | |
| *** amotoki has joined #openstack-neutron-release | 09:56 | |
| *** amotoki has quit IRC | 10:01 | |
| *** salv-orlando has joined #openstack-neutron-release | 10:03 | |
| *** amotoki has joined #openstack-neutron-release | 10:10 | |
| *** amotoki has quit IRC | 10:12 | |
| *** amotoki has joined #openstack-neutron-release | 10:13 | |
| *** ig0r_ has joined #openstack-neutron-release | 10:13 | |
| *** salv-orl_ has joined #openstack-neutron-release | 10:41 | |
| *** salv-orlando has quit IRC | 10:45 | |
| *** amotoki has quit IRC | 11:30 | |
| *** ig0r_ has quit IRC | 11:48 | |
| *** ig0r_ has joined #openstack-neutron-release | 11:49 | |
| *** yamamoto_ has quit IRC | 12:12 | |
| *** rtheis has joined #openstack-neutron-release | 12:25 | |
| *** salv-orl_ has quit IRC | 12:58 | |
| *** amotoki has joined #openstack-neutron-release | 13:33 | |
| *** amotoki has quit IRC | 13:43 | |
| *** amotoki has joined #openstack-neutron-release | 13:51 | |
| *** amotoki has quit IRC | 13:52 | |
| *** amotoki has joined #openstack-neutron-release | 13:58 | |
| *** rossella_s has quit IRC | 14:03 | |
| *** rossella_s has joined #openstack-neutron-release | 14:03 | |
| *** salv-orlando has joined #openstack-neutron-release | 14:13 | |
| *** amotoki has quit IRC | 14:45 | |
| *** amotoki has joined #openstack-neutron-release | 14:48 | |
| *** amotoki has quit IRC | 14:53 | |
| *** amotoki has joined #openstack-neutron-release | 15:09 | |
| *** yamamot__ has joined #openstack-neutron-release | 15:09 | |
| *** yamahata has joined #openstack-neutron-release | 15:27 | |
| *** iyamahat has joined #openstack-neutron-release | 15:27 | |
| *** salv-orlando has quit IRC | 16:01 | |
| *** ig0r_ has quit IRC | 16:20 | |
| *** armax has joined #openstack-neutron-release | 16:32 | |
| *** iyamahat has quit IRC | 16:53 | |
| *** yamamot__ has quit IRC | 16:59 | |
| *** yamamoto_ has joined #openstack-neutron-release | 17:00 | |
| *** yamahata has quit IRC | 17:10 | |
| *** yamahata has joined #openstack-neutron-release | 17:30 | |
| *** amotoki has quit IRC | 17:31 | |
| *** amotoki has joined #openstack-neutron-release | 18:14 | |
| *** salv-orlando has joined #openstack-neutron-release | 18:16 | |
| *** openstack has joined #openstack-neutron-release | 18:27 | |
| *** iyamahat has joined #openstack-neutron-release | 19:06 | |
| *** yamamoto_ has quit IRC | 19:34 | |
| *** yamamoto_ has joined #openstack-neutron-release | 20:48 | |
| *** yamamoto_ has quit IRC | 20:52 | |
| *** salv-orlando has quit IRC | 21:03 | |
| *** yamamoto_ has joined #openstack-neutron-release | 21:18 | |
| *** yamamoto_ has quit IRC | 21:22 | |
| *** yamamoto_ has joined #openstack-neutron-release | 21:48 | |
| *** yamamoto_ has quit IRC | 21:52 | |
| *** rossella_s has quit IRC | 22:03 | |
| *** rossella_s has joined #openstack-neutron-release | 22:03 | |
| *** rtheis has quit IRC | 22:30 | |
| *** yamamoto_ has joined #openstack-neutron-release | 22:48 | |
| *** yamamoto_ has quit IRC | 22:52 | |
| *** yamamoto_ has joined #openstack-neutron-release | 23:33 | |
| *** yamamoto_ has quit IRC | 23:37 | |
| *** salv-orlando has joined #openstack-neutron-release | 23:44 | |
| *** armax has quit IRC | 23:44 | |
| *** armax has joined #openstack-neutron-release | 23:44 | |
| *** salv-orlando has quit IRC | 23:53 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!