- openstack-discuss - lists.openstack.org

[dev][keystone] App Cred Capabilities Update
by Colleen Murphy 25 Feb '19

25 Feb '19

I have an initial draft of application credential capabilities available for review[1]. The spec[2] was not straightforward to implement. There were a few parts that I found unintuitive, user-unfriendly, and overcomplicated. The current proposed implementation differs from the spec in some ways that I want to discuss with the team. Given that non-client library freeze is next week, and the changes we need in keystonemiddleware also require changes in keystoneclient, I'm not sure there is enough time to properly flesh this out and allow for thorough code review, but if we miss the deadline we can be ready to get this in in the beginning of next cycle (with apologies to everyone waiting on this feature - I just want to make sure we get it right with minimal regrets). * Naming As always, naming is hard. In the spec, we've called the property that is attached to the app cred "capabilities" (for the sake of this email I'm going to call it user-created-rules), and we've called the operator-configured list of available endpoints "permissible path templates" (for the sake of this email I'm going to call it operator-created-rules). I find both confusing and awkward. "Permissible path templates" is not a great name because the rule is actually about more than just the path, it's about the request as a whole, including the method. I'd like to avoid saying "template" too because that evokes a picture of something like a Jinja or ERB template, which is not what this is, and because I'd like to avoid the whole string substitution thing - more on that below. In the implementation, I've renamed the operator-created-rules to "access rules". I stole this from Istio after Adam pointed out they have a really similar concept[3]. I really like this name because I think it well-describes the thing we're building without already being overloaded. So far, I've kept the user-created-rules as "capabilities" but I'm not a fan of it because it's an overloaded word and not very descriptive, although in my opinion it is still more descriptive than "whitelist" which is what we were originally going to call it. It might make sense to relate this property somehow to the operator-created-rules - perhaps by calling it access_rules_list, or granted_access_rules. Or we could call *this* thing the access rules, and call the other thing allowed_access_rules or permitted_access_rules. * Substitutions The way the spec lays out variable components of the URL paths for both user-created-rules and operator-created-rules is unnecessarily complex and in some cases faulty. The only way I can explain how complicated it is is to try to give an example: Let's say we want to allow a user to create an application credential that allows the holder to issue a GET request on the identity API that looks like /v3/projects/ef7284b4-3a75-4570-8ea8-b30214f18538/tags/foobar. The spec says that the string '/v3/projects/{project_id}/tags/{tag}' is what should be provided verbatim in the "path" attribute of a "capability", then there should be a "substitutions" attribute that sets {"tag": "foobar"}, then the project_id should be taken from the token scope at app cred usage time. When the capability is validated against the operator-created-rules at app cred creation time, it needs to check that the path string matches exactly, that the keys of the "substitutions" dict matches the "user template keys" list, and that keys required by the "context template keys" are provided by the token context. Taking the project ID, domain ID, or user ID from the token scope is not going to work because some of these APIs may actually be system-scoped APIs - it's just not a hard and fast rule that a project/domain/user ID in the URL maps to the same user and scope of the token used to create it. Once we do away with that, it stops making sense to have a separate attribute for the user-provided substitutions when they could just include that in the URL path to begin with. So the proposed implementation simply allows the wildcards * and ** in both the operator-created-rules and user-created-rules, no python-formatting variable substitutions. * UUIDs The spec says each operator-created-rule should have its own UUID, and the end user needs to look up and provide this UUID when they create the app cred rule. This has the benefit of having a fast lookup because we've put the onus on the user to look up the rule themselves, but I think it is very user-unfriendly. In the proposed implementation, I've done away with UUIDs on the operator-created-rules, and instead a match is looked up based on the service type, request path and method, and the "allow_chained" attribute (more on that next). Depending on how big we think this list of APIs will get, this will have some impact on performance on creation time (not at token validation time). UUIDs make sense for singleton resources that are created in a database. I imagine this starting as an operator-managed configuration file, and maybe some day in the future a catalog of this sort could be published by the services so that the operator doesn't have to maintain it themselves. To that end, I've implemented the operator-created-rules driver with a JSON file backend. But with this style of implementation, including UUIDs for every rule is awkward - it only makes sense if we're generating the resources within keystone and storing them in a database. * allow_chained The allow_chained attribute looks and feels awkward and adds complexity to the code Is there really a case when either the user or operator would not want to allow a service to make a request on behalf of a user who is making a more general request? Also, can we find a better name for it? Those are all the major question marks for me. There are some other minor differentiations from the spec and I will propose an update that makes it consistent with reality after we have these other questions sorted out. Colleen [1] https://review.openstack.org/#/q/topic:bp/whitelist-extension-for-app-creds [2] http://specs.openstack.org/openstack/keystone-specs/specs/keystone/stein/ca… [3] https://istio.io/docs/reference/config/authorization/istio.rbac.v1alpha1/#A…

2 6

[senlin] Forum session brainstorming
by Duc Truong 25 Feb '19

25 Feb '19

I have created an etherpad to capture ideas for Senlin forum sessions [1]. Please add your ideas along with your irc name. We especially would like to hear from Senlin users and operators on what you would like to see discussed. [1] https://etherpad.openstack.org/p/DEN-train-senlin-forum-brainstorming

1 0

UC Feb 2019 Election results
by Mohamed Elsakhawy 25 Feb '19

25 Feb '19

Good Afternoon all On behalf of the User Committee Elections officers, I am pleased to announce the results of the UC elections for Feb 2019. Please join me in congratulating the winners of the 3 seats : - Amy Marrich - Belmiro Moreira - John Studarus Thank you to all of the candidates and all of you who voted * https://civs.cs.cornell.edu/cgi-bin/results.pl?id=E_8760d5969c6275f1&rkey=7…

6 5

[dev][swift] Regarding fstab entries in a Swift controller
by Vedartham Bharath 25 Feb '19

25 Feb '19

Hi all, This is with regards to an issue in the documentation of Openstack Swift and mostly an issue to system administrators operating Swift.(Sorry if the subject tags are wrong!!) In the Swift Multiserver docs, When setting up an object storage server, the docs tell to use the disk labels in the disk's /etc/fstab entries. eg: /dev/sda /srv/node/sda xfs noatime............. I feel that we should encourage people to use UUIDs rather than disk labels. I have had a lot of issues with my Swift storage servers crashing whenever I reboot them. I have found out that the issue is with the /etc/fstab as the disk labels change whenever a disk is removed or added(depends on the OS's "mood" i.e boot order). I don't want to change my ring configuration every time I reboot my storage servers. Thank you Bharath

1 0

[openstack-dev] [Monasca] How to get “aggregated value of one metric statistics” ?
by Mohan Kumar 25 Feb '19

25 Feb '19

Hi Team, How to get “aggregated value of one metric statistics” from starting of month to till now . If I try to group metrics using * --period* based on timestamp it including data from previous month metrics as well [1] In below example , trying to get last ~24.5 days of metrics from particular tenant , But I can see 2019-01-26 data . With UTC_START_TIME “2019-02-01T00:00:00Z” [2] does *--merge_metrics * not Merge multiple metrics into a single result ? Please suggest how to customise my API call to get “AVG (aggregated) value of one metric statistics” from starting of month to till now . *Regards.,* Mohankumar N "*Confidentiality Warning*: This message and any attachments are intended only for the use of the intended recipient(s), are confidential and may be privileged. If you are not the intended recipient, you are hereby notified that any review, re-transmission, conversion to hard copy, copying, circulation or other use of this message and any attachments is strictly prohibited. If you are not the intended recipient, please notify the sender immediately by return email and delete this message and any attachments from your system. *Virus Warning:* Although the company has taken reasonable precautions to ensure no viruses are present in this email. The company cannot accept responsibility for any loss or damage arising from the use of this email or attachment."

2 2

[oslo] Feature freeze this week
by Ben Nemec 25 Feb '19

25 Feb '19

Just a reminder that Oslo feature freeze happens this week. Yes, it's earlier than everyone else, and if you're wondering why, we have a policy[1] that discusses it. The main thing is that if you have features you want to get into Oslo libraries this cycle, please make sure they merge by Friday. After that we'll need to go through the FFE process and there's no guarantee we can land them. Feel free to ping us on IRC if you need reviews. Thanks. -Ben 1: http://specs.openstack.org/openstack/oslo-specs/specs/policy/feature-freeze…

1 1

[tripleo] nominating Harald Jensås as a core reviewer
by Juan Antonio Osorio Robles 25 Feb '19

25 Feb '19

Hey folks! I would like to nominate Harald as a general TripleO core reviewer. He has consistently done quality reviews throughout our code base, helping us with great feedback and technical insight. While he has done a lot of work on the networking and baremetal sides of the deployment, he's also helped out on security, CI, and even on the tripleoclient side. Overall, I think he would be a great addition to the core team, and I trust his judgment on reviews. What do you think? Best regards

11 10

[openstack-dev] [Monasca] How to get “aggregated value of one metric statistics” ?
by Mohan Kumar 25 Feb '19

25 Feb '19

Hi Team, How to get “aggregated value of one metric statistics” from starting of month to till now . If I try to group metrics using * --period* based on timestamp it including data from previous month metrics as well [1] In below example , trying to get last ~24.5 days of metrics from particular tenant , But I can see 2019-01-26 data . With UTC_START_TIME “2019-02-01T00:00:00Z” [2] does *--merge_metrics * not Merge multiple metrics into a single result ? Please suggest how to customise my API call to get “AVG (aggregated) value of one metric statistics” from starting of month to till now . *Regards.,* Mohankumar N "*Confidentiality Warning*: This message and any attachments are intended only for the use of the intended recipient(s), are confidential and may be privileged. If you are not the intended recipient, you are hereby notified that any review, re-transmission, conversion to hard copy, copying, circulation or other use of this message and any attachments is strictly prohibited. If you are not the intended recipient, please notify the sender immediately by return email and delete this message and any attachments from your system. *Virus Warning:* Although the company has taken reasonable precautions to ensure no viruses are present in this email. The company cannot accept responsibility for any loss or damage arising from the use of this email or attachment."

1 0

[openstack-dev] [Monasca] How to get “aggregated value of one metric statistics” ?
by Mohan Kumar 25 Feb '19

25 Feb '19

Hi Team, How to get “aggregated value of one metric statistics” from starting of month to till now . If I try to group metrics using * --period* based on timestamp it including data from previous month metrics as well [1] In below example , trying to get last ~24.5 days of metrics from particular tenant , But I can see 2019-01-26 data . With UTC_START_TIME “2019-02-01T00:00:00Z” [2] does *--merge_metrics * not Merge multiple metrics into a single result ? Please suggest how to customise my API call to get “AVG (aggregated) value of one metric statistics” from starting of month to till now . *Regards.,* Mohankumar N "*Confidentiality Warning*: This message and any attachments are intended only for the use of the intended recipient(s), are confidential and may be privileged. If you are not the intended recipient, you are hereby notified that any review, re-transmission, conversion to hard copy, copying, circulation or other use of this message and any attachments is strictly prohibited. If you are not the intended recipient, please notify the sender immediately by return email and delete this message and any attachments from your system. *Virus Warning:* Although the company has taken reasonable precautions to ensure no viruses are present in this email. The company cannot accept responsibility for any loss or damage arising from the use of this email or attachment."

1 0

[searchlight] Team meeting cancelled today
by Trinh Nguyen 25 Feb '19

25 Feb '19

Hi team, I'll be out of the office for the rest of the day so I cannot hold the team meeting at 13:30 UTC today. Meeting will happen again on 11th March. Bests, -- *Trinh Nguyen* *www.edlab.xyz <https://www.edlab.xyz>*

1 0