| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2018-05-04 | |||
| 13:50:55 | mordred | Shrews: looking | |
| 13:51:21 | mordred | Shrews: also the sdk functional tests keep hitting flapping errors related to floating ips - so I'm trying to clean that up | |
| 13:51:36 | Shrews | yay | |
| 13:52:33 | openstackgerrit | David Shrewsbury proposed openstack/os-client-config master: Add cache API methods back to OCC https://review.openstack.org/566308 | |
| 13:52:35 | frickler | Shrews: hmm, I think you are patching the wrong location, the issue for me is with config.OpenStackConfig() and not cloud_config.Config() | |
| 13:53:03 | mordred | AH | |
| 13:53:04 | Shrews | pooja_jadhav: ah, Closes-Bug. couldn't remember it and couldn't find it any docs (it's all shuffled since last i looked) | |
| 13:53:07 | mordred | that makes more sense to me | |
| 13:53:14 | frickler | Shrews: also yes, would be good to have tests for that | |
| 13:55:06 | pooja_jadhav | Shrews: Hi, I am not getting what you are trying to say | |
| 13:55:21 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Clean up floating ip tests https://review.openstack.org/566320 | |
| 13:55:58 | Shrews | pooja_jadhav: ooh, sorry. that was meant for someone else | |
| 13:56:38 | pooja_jadhav | shrews: ohk | |
| 13:57:29 | openstackgerrit | Merged openstack/os-client-config master: fix link to openstacksdk in deprecation block https://review.openstack.org/566318 | |
| 13:58:07 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: close files after open in unit/base https://review.openstack.org/566321 | |
| 13:58:07 | Shrews | ok, i'm confused since those calls originally existed in CloudConfig | |
| 13:59:45 | frickler | Shrews: this is the code sample I posted on the bug report which is broken with the new release https://github.com/ansible/ansible/blob/devel/contrib/inventory/openstack.py#L193-L204 | |
| 14:00:03 | openstackgerrit | Monty Taylor proposed openstack/os-client-config master: Allow monkeypatching CONFIG_FILES https://review.openstack.org/566139 | |
| 14:00:04 | openstackgerrit | Monty Taylor proposed openstack/os-client-config master: Add cache methods back to OpenSackConfig https://review.openstack.org/566322 | |
| 14:00:12 | mordred | Shrews: there were two sets of them | |
| 14:00:28 | Shrews | mordred: oh. well that's not confusing | |
| 14:00:34 | mordred | yah - it was terrible | |
| 14:00:37 | mordred | the CloudConfig ones originally called the OpenStackConfig ones | |
| 14:01:32 | mordred | BUT - if you're just working with CloudConfig objects that's awkward - so CloudConfig got fixed to not need the OpenStackConfig methods anymore - and in sdk OpenStackConfig doens't have the methods because they're silly | |
| 14:01:53 | mordred | and I totally missed keeping them for backwards compat in the occ patch | |
| 14:02:27 | Shrews | i'm not clear on the fix then | |
| 14:02:47 | mordred | adding them back to OpenStackConfig so that occ users who were using them don't break | |
| 14:10:25 | Shrews | mordred: how do you get from occ.OpenStackConfig to the osdk.CloudRegion where those calls are defined? | |
| 14:10:49 | Shrews | those calls seem like they should be region independent | |
| 14:11:19 | Shrews | or is it allowed to have different cache characteristics per region? | |
| 14:12:00 | mordred | no, you're right- it's region independent | |
| 14:12:14 | mordred | thing is - the CloudRegion is the primary thing people work with | |
| 14:14:06 | mordred | so - basically, the original change was to have the cache settings be actual constructor arguments to CloudRegion rather than having to pass an OpenStackConfig to CloudRegion | |
| 14:14:30 | mordred | get_one_cloud then started passing the values to CloudRegion when it created one | |
| 14:14:53 | dmsimard | mordred: how would you do a search_ports for a fixed_ip ? | |
| 14:15:22 | mordred | dmsimard: conn.search_ports(filters={'fixed_ip': blah}) I believe | |
| 14:15:29 | dmsimard | http://paste.openstack.org/show/720377/ doesn't seem to work | |
| 14:15:35 | dmsimard | mordred: it's nested under fixed_ips | |
| 14:15:53 | dmsimard | mordred: i.e, http://paste.openstack.org/show/720378/ | |
| 14:16:15 | mordred | of course it is | |
| 14:16:31 | mordred | dmsimard: let's step back one - what are you trying to accomplish? | |
| 14:17:01 | dmsimard | mordred: are you sitting on something ? | |
| 14:17:15 | dmsimard | mordred: I need to add all the IPs from a subnet to another subnet's ports allowed-address-pairs | |
| 14:17:30 | dmsimard | basically make the ports of a subnet able to bind the ips from another subnet | |
| 14:17:39 | mordred | wow | |
| 14:17:42 | mordred | that's awesome | |
| 14:17:53 | dmsimard | It's not hard, just need to figure that one thing | |
| 14:18:54 | mordred | dmsimard: so - you have two choices ... | |
| 14:19:06 | dmsimard | https://github.com/openstack-infra/shade/blob/16e7290ff45ca6fa5d7ba7360575c38e0b0fd043/shade/openstackcloud.py#L6237-L6247 :/ | |
| 14:19:37 | Shrews | shade only supports filtering dictionaries of dictionaries. lists of dictionaries are not supported | |
| 14:19:42 | mordred | dmsimard: you can pass a jmsepath string to filters - or you can do a list ports and filter yourself | |
| 14:19:42 | Shrews | yeah, that | |
| 14:20:10 | mordred | listing ports and filtering yourself is likely the easier thing | |
| 14:20:22 | Shrews | i forgot we added jmespath support | |
| 14:20:35 | mordred | but if you wanna play with the jmsepath supporT: http://jmespath.org/ | |
| 14:20:52 | dmsimard | mordred: I can filter myself I suppose | |
| 14:22:00 | mordred | Shrews: how about this ... | |
| 14:22:31 | Shrews | mordred: if the fix is clear in your head, feel free to add a PS to my review | |
| 14:22:53 | Shrews | i apparently need more coffee to see it | |
| 14:23:13 | mordred | Shrews: oh - I already pushed up https://review.openstack.org/566322 (should have done it as a PS to yours) | |
| 14:24:13 | mordred | Shrews: that said - I'm thinking maybe we just revert and release that - then add some more tests/test jobs and try the ectomy again | |
| 14:24:26 | mordred | Shrews: since we're clearly missing some coverage on ansible impact of occ changes | |
| 14:24:42 | dmsimard | mordred: the thing is I need to create a port if there's no port for that ip address. Do you have an idea other than iterating through the ports and fixed_ips and creating them if they don't exist ? | |
| 14:24:58 | Shrews | mordred: possibly a good idea | |
| 14:25:15 | mordred | dmsimard: nope. I recommend brute force | |
| 14:25:38 | dmsimard | mordred: I wonder if I could blindly fire create_ports and basically pass on exception | |
| 14:25:44 | dmsimard | but that'd be kind of dirty | |
| 14:25:49 | Shrews | mordred: i'll abandon mine assuming the revert | |
| 14:26:15 | dmsimard | mordred: hmmm.. how does http://docs.ansible.com/ansible/latest/modules/os_port_module.html handle idempotency ? | |
| 14:27:18 | openstackgerrit | Monty Taylor proposed openstack/os-client-config master: Revert "Replace guts with openstack.config" https://review.openstack.org/566328 | |
| 14:27:29 | mordred | Shrews: ^^ | |
| 14:27:58 | Shrews | mordred: is that going to again break nodepool now that https://review.openstack.org/566138 *just* merged? | |
| 14:29:31 | dmsimard | mordred: looks like os_port basically relies on the name of the port.. which is not stupid | |
| 14:29:42 | mordred | Shrews: ugh. probably. we should get the followup patch fixed | |
| 14:29:57 | pooja_jadhav | mordred: Hi | |
| 14:31:02 | mordred | Shrews: maybe we should try the fix patch instead of the revert | |
| 14:32:45 | Shrews | mordred: your patch can't work i think. there is no self.get_cache_expiration_time() | |
| 14:33:09 | Shrews | mordred: that's only in CloudRegion, which is why i asked how to get to that | |
| 14:34:42 | Shrews | as are the _cache_* attrs | |
| 14:35:09 | Shrews | oh, no | |
| 14:35:13 | Shrews | wrong about the attrs | |
| 14:35:24 | mordred | the patch adds get_cache_expiration_time | |
| 14:35:48 | Shrews | oh duh | |
| 14:36:10 | Shrews | *sigh | |
| 14:36:36 | Shrews | mordred: ok, let's just try your patch first | |
| 14:37:06 | mordred | Shrews: yah - this *is* accessing private members of the osdk base class ... so we'll need to add some tests to occ (which we were obviously missing anyway) to make sure those methods continue to work | |
| 14:37:25 | mordred | we run occ tests on sdk patches, so that should prevent sdk patches from changing those private members | |
| 14:37:37 | mordred | and yes, that's terrible software engineering practice in general | |
| 14:40:57 | openstackgerrit | Monty Taylor proposed openstack/os-client-config master: Revert "Replace guts with openstack.config" https://review.openstack.org/566328 | |
| 14:41:12 | mordred | Shrews: well, there's that (without merge conflict this time) in case we decide we want to go that route | |
| 14:46:08 | Shrews | mordred: i think the nodepool followup to use osdk is failing because osdk defines a get_cache_expirations() but shade is calling get_cache_expiration() | |
| 14:46:23 | Shrews | mordred: accidental pluralization there? | |
| 14:47:45 | Shrews | http://logs.openstack.org/58/566158/1/check/nodepool-functional-py35/45d52ed/controller/logs/screen-nodepool-builder.txt.gz#_May_04_05_56_28_537111 | |
| 14:48:13 | mordred | Shrews: crap | |
| 14:48:22 | mordred | Shrews: the plural/singular is intentional | |
| 14:48:24 | mordred | but | |
| 14:48:44 | mordred | not particularly awesome for this transition | |
| 14:51:32 | Shrews | so i think pushing your occ patch, given that ^ which would require a new osdk release, is probably the thing to do now? | |
| 14:51:46 | mordred | yah | |
| 14:51:59 | mordred | and I'll also work on a shade patch to use osdk for config directly | |