| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2022-10-04 | |||
| 12:23:27 | gtema | the place you mention is exactly one of the black holes | |
| 12:23:31 | jm1 | gtema: we also have subnets | |
| 12:23:56 | jm1 | gtema: the issue is, that this line here might return the "wrong" network https://opendev.org/openstack/openstacksdk/src/commit/093e71e5a6050e6f7d47a03b5f03a62b8f60ca60/openstack/cloud/_floating_ip.py#L1109 | |
| 12:24:13 | jm1 | gtema: it might return the one the server is connected to, but it could also return the other one | |
| 12:24:42 | gtema | because historically it searches for this "NAT network" which should be a private net server is connected to and connected through router to public net | |
| 12:25:06 | gtema | do you have logs (or how to reproduce? | |
| 12:25:15 | jm1 | gtema: https://6aef0912e1e7e067f94c-3957ec525977d9a3771da3eb17c87a1d.ssl.cf5.rackcdn.com/852119/18/check/ansible-collections-openstack-functional-devstack/593fd7f/job-output.txt | |
| 12:26:03 | jm1 | gtema: here is the culprit: https://opendev.org/openstack/openstacksdk/src/commit/093e71e5a6050e6f7d47a03b5f03a62b8f60ca60/openstack/cloud/_network_common.py#L180 | |
| 12:26:34 | jm1 | gtema: this will look though all subnets which will be returned even though the server might not be attached to that | |
| 12:26:47 | gtema | which which change it is (so that I reconstruct conditions)? | |
| 12:27:12 | jm1 | gtema: https://review.opendev.org/c/openstack/ansible-collections-openstack/+/852119/18 | |
| 12:27:59 | jm1 | gtema: my approach would be to return all possible nat destinations and let the caller choose what the best is | |
| 12:28:34 | gtema | actually there is clouds.yaml parameter for manual name of such net | |
| 12:29:24 | jm1 | gtema: but _nat_destination_port() can decide what nat destination to use without the user having to help the sdk | |
| 12:30:05 | jm1 | gtema: _nat_destination_port gets a server and has a list of ports. it just has to check which port is actually on a nat_destination network | |
| 12:30:05 | gtema | I told you this is now all quite historical and not too much scalable, haven't I? | |
| 12:30:13 | jm1 | gtema: hmm | |
| 12:30:16 | jm1 | gtema: so ignore? | |
| 12:30:37 | gtema | nope, I want to have a functest for that directly in SDK | |
| 12:31:27 | jm1 | gtema: the use case is pretty straight forward though: one has a server and wants it to have a floating ip. this should not fail just because the user has two networks with a gateway.. | |
| 12:31:30 | gtema | https://review.opendev.org/q/topic:better_tests is a chain where this would need to go | |
| 12:31:56 | gtema | this is clear, jm1. But I would need to create a proper functest for that | |
| 12:32:36 | jm1 | gtema: sure. question is, should i write a patch for this _nat_destination_port or is this going away anyway? | |
| 12:33:14 | gtema | no, I would first want to have a test that reproduces error and with that start reworking of the whole madness | |
| 12:34:55 | jm1 | gtema: ok sooo.. should i write a test? | |
| 12:35:13 | gtema | just describe to me detailed conditions | |
| 12:36:03 | jm1 | gtema: here or in a bug report? | |
| 12:36:29 | gtema | doesn't matter - even an email to me. Important is that I do not loose it | |
| 12:36:47 | gtema | and bugreport is exactly the place to loose everything once you close the tab | |
| 12:44:20 | jm1 | gtema: ok. devstack default setup with public network. 2 private networks with a subnet each. all three networks attached to the router. server attached to one private network. call to add_ips_to_server(server, auto_ip=True, reuse=False). sometimes fails with "No port on server... was found matching your NAT destination network" because _nat_destination_port() returns the other private network, not the one the server has been attached to. | |
| 12:44:29 | jm1 | gtema: plus links above. enough? | |
| 12:46:23 | gtema | oh, so actually all private nets attached to one router? that is a great fact to know | |
| 12:49:56 | jm1 | gtema: yes. although this would not change the bug because it lists all subnets anyway | |
| 12:51:02 | jm1 | gtema: have more than one network with a gateway? boom | |
| 12:51:21 | jm1 | gtema: or rather "potential boom" | |
| 12:52:03 | gtema | where is the connectivity between pub net and router? don't see it in the change | |
| 12:53:10 | jm1 | gtema: https://review.opendev.org/c/openstack/ansible-collections-openstack/+/852119/18/ci/roles/server/tasks/main.yml#37 | |
| 12:53:24 | gtema | ah, overseen | |
| 12:53:47 | gtema | ok, will sketch the test | |
| 12:56:07 | jm1 | gtema: while you are at it.. https://storyboard.openstack.org/#!/story/2010153 ;) | |
| 12:56:40 | jm1 | gtema: he also submitted a patch | |
| 12:57:26 | gtema | ugh, I don't know how to deal with patches into something that is so fragile | |
| 12:59:07 | jm1 | gtema: this patch actually looks good. it does not remove or change functionality, it just adds a missing parameter | |
| 12:59:41 | gtema | which works "magically" | |
| 13:00:06 | gtema | there are no really reasonable tests for this whole "chapter" | |
| 13:00:19 | jm1 | gtema: so you cannot break something with merging this | |
| 13:00:34 | gtema | lol - I can break you even further | |
| 13:02:09 | jm1 | gtema: not with this patch though ;) | |
| 13:02:28 | gtema | sure? | |
| 13:03:11 | jm1 | gtema: not more than what is broken already | |
| 13:03:13 | gtema | it doesn't look bad, I am just too skeptic since there are no real tests here | |
| 13:03:35 | gtema | okay, then I can go with merging it and restructure with it being added | |
| 13:06:12 | opendevreview | Artem Goncharov proposed openstack/openstacksdk master: support nat_destination when attaching existing floating_ip to a server https://review.opendev.org/c/openstack/openstacksdk/+/850115 | |
| 13:18:53 | jm1 | gtema: i know you hate storyboard but i need a place to track that error https://storyboard.openstack.org/#!/story/2010352 | |
| 13:19:25 | gtema | I do not hate it - I do not use it ;-) | |
| 13:19:41 | gtema | because it does not help, but rather disturb | |
| 13:45:46 | gtema | really? okay, then I just don't push any other big changes ;-) | |
| 13:56:48 | opendevreview | Merged openstack/ansible-collections-openstack stable/1.0.0: Release 1.10.0 version https://review.opendev.org/c/openstack/ansible-collections-openstack/+/860223 | |
| 14:35:46 | gtema | you know jm1, while now reworking tests to be working in regular clouds faced already something fun errors ("Multiple possible networks found, use a Network ID to be more specific.") | |
| 14:39:03 | jm1 | gtema: i am feeling you 😬 | |
| 15:19:35 | opendevreview | Artem Goncharov proposed openstack/openstacksdk master: Whitelist cloud functional tests in acceptance https://review.opendev.org/c/openstack/openstacksdk/+/860009 | |
| 18:15:38 | opendevreview | Merged openstack/cliff master: Replace abc.abstractproperty with property and abc.abstractmethod https://review.opendev.org/c/openstack/cliff/+/852058 | |
| 18:15:40 | opendevreview | Merged openstack/cliff master: columns: Useful __str__, __repr__ implementation https://review.opendev.org/c/openstack/cliff/+/858550 | |
| 19:05:17 | opendevreview | Jakob Meng proposed openstack/ansible-collections-openstack master: Replaced expensive get_server() and fixed issues in server module https://review.opendev.org/c/openstack/ansible-collections-openstack/+/852119 | |
| #openstack-sdks - 2022-10-05 | |||
| 05:49:45 | opendevreview | Dr. Jens Harbott proposed openstack/python-openstackclient master: DNM: Test job with capped sdk https://review.opendev.org/c/openstack/python-openstackclient/+/858569 | |
| 08:52:08 | opendevreview | Jakob Meng proposed openstack/ansible-collections-openstack master: Replaced expensive get_server() and fixed issues in server module https://review.opendev.org/c/openstack/ansible-collections-openstack/+/852119 | |
| 08:52:35 | opendevreview | Jakob Meng proposed openstack/ansible-collections-openstack master: Replaced expensive get_server() and fixed issues in server module https://review.opendev.org/c/openstack/ansible-collections-openstack/+/852119 | |
| 08:59:57 | opendevreview | Jakob Meng proposed openstack/openstacksdk master: Added Ansible OpenStack Collection to Bifrost's job.required-projects https://review.opendev.org/c/openstack/openstacksdk/+/860351 | |
| 09:02:20 | opendevreview | Jakob Meng proposed openstack/ansible-collections-openstack master: Updates server_volume for 2.0.0 https://review.opendev.org/c/openstack/ansible-collections-openstack/+/858834 | |
| 10:02:24 | opendevreview | Jakob Meng proposed openstack/ansible-collections-openstack master: Refactor server_volume to be compatible with openstacksdk>=0.99.0 https://review.opendev.org/c/openstack/ansible-collections-openstack/+/858834 | |
| 10:27:20 | opendevreview | Artem Goncharov proposed openstack/openstacksdk master: Implement acceptance test job https://review.opendev.org/c/openstack/openstacksdk/+/860399 | |
| 10:30:54 | opendevreview | Artem Goncharov proposed openstack/openstacksdk master: Implement acceptance test job https://review.opendev.org/c/openstack/openstacksdk/+/860399 | |
| 10:45:16 | opendevreview | Artem Goncharov proposed openstack/openstacksdk master: Implement acceptance test job https://review.opendev.org/c/openstack/openstacksdk/+/860399 | |
| 11:10:30 | opendevreview | Artem Goncharov proposed openstack/openstacksdk master: Implement acceptance test job https://review.opendev.org/c/openstack/openstacksdk/+/860399 | |
| 11:51:26 | opendevreview | Artem Goncharov proposed openstack/openstacksdk master: Implement acceptance test job https://review.opendev.org/c/openstack/openstacksdk/+/860399 | |
| 12:51:19 | opendevreview | Artem Goncharov proposed openstack/openstacksdk master: Implement acceptance test job https://review.opendev.org/c/openstack/openstacksdk/+/860399 | |
| 13:40:20 | opendevreview | Artem Goncharov proposed openstack/openstacksdk master: Implement acceptance test job https://review.opendev.org/c/openstack/openstacksdk/+/860399 | |
| 13:51:36 | jm1 | gtema: can we ask sdk for availability of block_storage and network endpoints? | |
| 13:51:47 | gtema | yes | |
| 13:52:16 | gtema | https://opendev.org/openstack/openstacksdk/src/branch/master/openstack/tests/functional/block_storage/v3/base.py | |
| 13:52:48 | gtema | there are, however, as found during functesting, further fun facts of having services but disabled | |
| 13:53:01 | gtema | anyway - that one is the most useful query | |
| 13:53:33 | jm1 | gtema: oh great, thank you! | |
| 13:53:40 | gtema | wlcm | |
| 13:53:52 | jm1 | gtema: i was grep'ing for endpoint but it was has_service 😅 | |
| 13:54:44 | gtema | there is also `self.conn._has_neutron_extension('fwaas_v2')`for details about neutron extensions availability | |
| 13:58:21 | jm1 | gtema: we only want to ask for availability of block-storage and network and are assume compute to be available. we need that for quota module. yeah i know, not optimal, but we postpone a proper split of the quota module for later, no time atm | |
| 13:59:06 | gtema | quota is a bit a pain - https://review.opendev.org/c/openstack/openstacksdk/+/860009/3/openstack/tests/functional/cloud/test_quotas.py | |
| 13:59:35 | gtema | so - yes, right | |
| 13:59:50 | gtema | actually the only place where network is not going to be present is RAX | |
| 14:00:24 | jm1 | gtema: wait, what is volume service? | |
| 14:00:36 | gtema | hehe, this is fun | |
| 14:00:53 | gtema | this is a synonym for block-storage | |
| 14:01:07 | jm1 | gtema: oh ok, dont shock me again please.. | |
| 14:01:09 | gtema | since in the service catalog it is actually volume/volumev2/volumev3 | |
| 14:01:20 | gtema | https://opendev.org/openstack/os-service-types/src/branch/master/os_service_types/data/service-types.json#L35 | |
| 14:01:48 | jm1 | gtema: okay... i appreciate that you are working on sdk 😅 | |
| 14:01:54 | jm1 | gtema: ..not me... | |
| 14:02:20 | gtema | I do, I try, I suffer | |
| 14:02:23 | jm1 | gtema: block-store... can we please add volume-store and volume-storage and block-storage-v3 and .. 🙊 | |