Re: [openstack-dev] [forum] Future of Stackalytics
On 2017-06-12 14:07:51 -0700 (-0700), Ken'ichi Ohmichi wrote: [...] > The difference between current stackalytics config and the above API > is stackalytics contains gerrit-id and launchpad-id on the config > but the API doesn't. I guess we can use e-mail address instead of > gerrit-id and launchpad-id and hope them from stackalytics config. [...] Right, I expect a sane design to be using E-mail addresses to identify the mapping between the foundation's data, Gerrit and Launchpad (this would also make it possible to correlate mailing list posts if we wanted). Something along the lines of querying Gerrit for the list of known E-mail addresses and then querying the foundation directory for any profiles associated with each of those addresses, building up a set of unique foundation IDs as it goes. Once authentication for Gerrit moves to OpenStackID we can more directly correlate Gerrit users and foundation profiles based on a common OpenID. -- Jeremy Stanley signature.asc Description: Digital signature __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
2017-06-08 10:51 GMT-07:00 Jeremy Stanley: > On 2017-06-08 09:49:03 -0700 (-0700), Ken'ichi Ohmichi wrote: >> 2017-06-08 7:19 GMT-07:00 Jeremy Stanley : > [...] >> > There is a foundation member directory API now which provides >> > affiliation details and history, so if it were my project (it's >> > not though) I'd switch to querying that and delete all the >> > static affiliation mapping out of that config instead. Not only >> > would it significantly reduce the reviewer load for >> > Stackalytics, but it would also provide a greater incentive for >> > contributors to keep their affiliation data updated in the >> > foundation member directory. >> >> Interesting idea, thanks. It would be nice to centralize such >> information into a single place. Can I know the detail of the API? >> I'd like to take a look for some prototyping. > > It only _just_ rolled to production at > https://openstackid-resources.openstack.org/api/public/v1/members > yesterday so I don't know how stable it should be considered at this > particular moment. The implementation is at > https://git.openstack.org/cgit/openstack-infra/openstackid-resources/tree/app/Models/Foundation/Main/Member.php > > > but details haven't been added to the API documentation in that repo > yet. (I also just now realized we haven't added a publishing job for > those API docs either, so I'm working on that bit immediately.) > > The relevant GET parameters for this case are > filter=email==someb...@example.com and relations=all_affiliations > which gets you a list under the "affiliations" key with all > start/end dates and organizations for the member associated with > that address. This of course presumes contributors update their > foundation profiles to include any E-mail addresses they use with > Git, as well as recording appropriate affiliation timeframes. Those > fields in the member directory profiles have existed for quite a few > years now, so hopefully at least some of us have already done that. Thanks for the info, Jeremy. The difference between current stackalytics config and the above API is stackalytics contains gerrit-id and launchpad-id on the config but the API doesn't. I guess we can use e-mail address instead of gerrit-id and launchpad-id and hope them from stackalytics config. I will dig more deeply anyways. Thanks __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
On 06/11/2017 03:17 PM, Matt Riedemann wrote: On 6/8/2017 12:57 AM, Adam Harwell wrote: As a core reviewer for LBaaS I actually find Stackalytics quite helpful for giving me a quick snapshot of contributions, and it lines up almost perfectly in my experience with what I see when I'm actually reviewing and working with people (if you know which statistics to look at -- just sorting by sheer number of reviews or commits and ignoring everything else is of course not useful, and as you say possibly misleading). In all though I actually find that it is a very accurate representation of people's work. For example, in looking at reviewer contributions, I make a mental score based on both the number of reviews, but also the +% (this shouldn't be too high) and the disagreement score (low is generally good, but 0% with a high review count might be questionable). So, I know to discount someone who just spams +1 at everything that has a +2 already and doesn't contribute anything else, which can go unnoticed while reading reviews but sticks out like a sore thumb in Stackalytics. The other side of the coin is someone who posts a ton of useless comments and -1's everything, which then is super obvious to anyone who actually reads reviews. Maybe the experience with the projects I work on is a little different than some of the more populous "base" services like Nova or Neutron? Regardless, I'd be really sad to see it go, as I use it multiple times a week for various reasons. So, I definitely agree with keeping it around and possibly focusing on improving the way the data is displayed. It is definitely best used as one tool in a toolkit, not taken alone as a single source of truth. Is that the main problem people are trying to solve? I agree with you, and the experience in Nova is the same. I use Stackalytics the same way. Note, however, that reviewstats is also published from a site that russellb has running, e.g.: http://russellbryant.net/openstack-stats/nova-reviewers-30.txt I sometimes use Stackalytics to get a list of person's latest reviews (e.g. [1]) or limit to to a certain project (e.g. [2]). There is something called "Person-day effort" (e.g. [3]), which looks also interesting, but I haven't looked into it. [1] http://stackalytics.com/?user_id=divius [2] http://stackalytics.com/?user_id=divius=ironic-group [3] http://stackalytics.com/?user_id=divius=ironic-group=person-day __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
On 6/8/2017 12:57 AM, Adam Harwell wrote: As a core reviewer for LBaaS I actually find Stackalytics quite helpful for giving me a quick snapshot of contributions, and it lines up almost perfectly in my experience with what I see when I'm actually reviewing and working with people (if you know which statistics to look at -- just sorting by sheer number of reviews or commits and ignoring everything else is of course not useful, and as you say possibly misleading). In all though I actually find that it is a very accurate representation of people's work. For example, in looking at reviewer contributions, I make a mental score based on both the number of reviews, but also the +% (this shouldn't be too high) and the disagreement score (low is generally good, but 0% with a high review count might be questionable). So, I know to discount someone who just spams +1 at everything that has a +2 already and doesn't contribute anything else, which can go unnoticed while reading reviews but sticks out like a sore thumb in Stackalytics. The other side of the coin is someone who posts a ton of useless comments and -1's everything, which then is super obvious to anyone who actually reads reviews. Maybe the experience with the projects I work on is a little different than some of the more populous "base" services like Nova or Neutron? Regardless, I'd be really sad to see it go, as I use it multiple times a week for various reasons. So, I definitely agree with keeping it around and possibly focusing on improving the way the data is displayed. It is definitely best used as one tool in a toolkit, not taken alone as a single source of truth. Is that the main problem people are trying to solve? I agree with you, and the experience in Nova is the same. I use Stackalytics the same way. Note, however, that reviewstats is also published from a site that russellb has running, e.g.: http://russellbryant.net/openstack-stats/nova-reviewers-30.txt -- Thanks, Matt __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
On June 8, 2017 at 07:20:23, Jeremy Stanley (fu...@yuggoth.org) wrote: > > On 2017-06-07 16:36:45 > > -0700(http://airmail.calendar/2017-06-07%2016:36:45%20PDT) > (-0700), Ken'ichi Ohmichi wrote: > [...] > > one of config files is 30KL due to much user information and that > > makes the maintenance hard now. I am trying to separate user > part > > from the existing file but I cannot find the way to make a > > consensus for such thing. > > There is a foundation member directory API now which provides > affiliation details and history, so if it were my project (it's > not > though) I'd switch to querying that and delete all the static > affiliation mapping out of that config instead. Not only would > it > significantly reduce the reviewer load for Stackalytics, but > it > would also provide a greater incentive for contributors to keep > their affiliation data updated in the foundation member directory. +1. This would really help me when generating the stats for our yearly reports/keynote/etc stats instead of having to query multiple sources and figure out which one is more current. — Mike Perez __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
On 2017-06-08 09:49:03 -0700 (-0700), Ken'ichi Ohmichi wrote: > 2017-06-08 7:19 GMT-07:00 Jeremy Stanley: [...] > > There is a foundation member directory API now which provides > > affiliation details and history, so if it were my project (it's > > not though) I'd switch to querying that and delete all the > > static affiliation mapping out of that config instead. Not only > > would it significantly reduce the reviewer load for > > Stackalytics, but it would also provide a greater incentive for > > contributors to keep their affiliation data updated in the > > foundation member directory. > > Interesting idea, thanks. It would be nice to centralize such > information into a single place. Can I know the detail of the API? > I'd like to take a look for some prototyping. It only _just_ rolled to production at https://openstackid-resources.openstack.org/api/public/v1/members yesterday so I don't know how stable it should be considered at this particular moment. The implementation is at https://git.openstack.org/cgit/openstack-infra/openstackid-resources/tree/app/Models/Foundation/Main/Member.php > but details haven't been added to the API documentation in that repo yet. (I also just now realized we haven't added a publishing job for those API docs either, so I'm working on that bit immediately.) The relevant GET parameters for this case are filter=email==someb...@example.com and relations=all_affiliations which gets you a list under the "affiliations" key with all start/end dates and organizations for the member associated with that address. This of course presumes contributors update their foundation profiles to include any E-mail addresses they use with Git, as well as recording appropriate affiliation timeframes. Those fields in the member directory profiles have existed for quite a few years now, so hopefully at least some of us have already done that. -- Jeremy Stanley signature.asc Description: Digital signature __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
2017-06-08 7:19 GMT-07:00 Jeremy Stanley: > On 2017-06-07 16:36:45 -0700 (-0700), Ken'ichi Ohmichi wrote: > [...] >> one of config files is 30KL due to much user information and that >> makes the maintenance hard now. I am trying to separate user part >> from the existing file but I cannot find the way to make a >> consensus for such thing. > > There is a foundation member directory API now which provides > affiliation details and history, so if it were my project (it's not > though) I'd switch to querying that and delete all the static > affiliation mapping out of that config instead. Not only would it > significantly reduce the reviewer load for Stackalytics, but it > would also provide a greater incentive for contributors to keep > their affiliation data updated in the foundation member directory. Interesting idea, thanks. It would be nice to centralize such information into a single place. Can I know the detail of the API? I'd like to take a look for some prototyping. >> In addition, we have two ways for managing bug reports: launchpad and >> storyboard if considering it as infra project. > > It's not (at least presently) an Infrastructure team deliverable. > It's only an unofficial project which happens to have granted the > infra-core team approval rights (for reasons I don't recall, if I > ever even knew it was the case before now). > >> It would be necessary to transport them from launchpad, I guess. > [...] > > If its maintainers want to migrate from LP to SB, we already have an > import script which copies in all the existing bug reports so that's > not really a challenge. Oh, cool. Glad to hear that :) Thanks __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
On Thu, Jun 08, 2017 at 02:11:18PM +, Jeremy Stanley wrote: > On 2017-06-08 09:22:48 -0400 (-0400), Paul Belanger wrote: > [...] > > We also have another issue where we loose access to gerrit and our apache > > process pins CPU to 100%, these might also be low hanging friut for people > > wanting to get involved. > [...] > > Wasn't that fixed with a new lower-bound on paramiko so it now > closes SSH API sessions correctly? > We did submit a patch, but I believe we are still leaking some connections to gerrit. We likley need to audit the code to ensure we applied the patch to all connection attempts. __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
On 2017-06-07 16:36:45 -0700 (-0700), Ken'ichi Ohmichi wrote: [...] > one of config files is 30KL due to much user information and that > makes the maintenance hard now. I am trying to separate user part > from the existing file but I cannot find the way to make a > consensus for such thing. There is a foundation member directory API now which provides affiliation details and history, so if it were my project (it's not though) I'd switch to querying that and delete all the static affiliation mapping out of that config instead. Not only would it significantly reduce the reviewer load for Stackalytics, but it would also provide a greater incentive for contributors to keep their affiliation data updated in the foundation member directory. > In addition, we have two ways for managing bug reports: launchpad and > storyboard if considering it as infra project. It's not (at least presently) an Infrastructure team deliverable. It's only an unofficial project which happens to have granted the infra-core team approval rights (for reasons I don't recall, if I ever even knew it was the case before now). > It would be necessary to transport them from launchpad, I guess. [...] If its maintainers want to migrate from LP to SB, we already have an import script which copies in all the existing bug reports so that's not really a challenge. -- Jeremy Stanley signature.asc Description: Digital signature __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
On 2017-06-08 09:22:48 -0400 (-0400), Paul Belanger wrote: [...] > We also have another issue where we loose access to gerrit and our apache > process pins CPU to 100%, these might also be low hanging friut for people > wanting to get involved. [...] Wasn't that fixed with a new lower-bound on paramiko so it now closes SSH API sessions correctly? -- Jeremy Stanley __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
On Wed, May 17, 2017 at 06:55:57PM +, Jeremy Stanley wrote: > On 2017-05-17 16:16:30 +0200 (+0200), Thierry Carrez wrote: > [...] > > we need help with completing the migration to infra. If interested > > you can reach out to fungi (Infra team PTL) nor mrmartin (who > > currently helps with the transition work). > [...] > > The main blocker for us right now is addressed by an Infra spec > (Stackalytics is an unofficial project and it's unclear to us where > design discussions for it happen): > > https://review.openstack.org/434951 > > In particular, getting the current Stackalytics developers on-board > with things like this is where we've been failing to make progress > mainly (I think) because we don't have a clear venue for discussions > and they're stretched pretty thin with other work. If we can get > some additional core reviewers for that project (and maybe even talk > about turning it into an official team or joining them up as a > deliverable for an existing team) that might help. > -- > Jeremy Stanley Agree with Jeremy. We have been running a shadow instances of stackalytics[1] for 2 years now. So, we could make the flip today to our community infrastructure if we wanted too. The persistent cache would be a helpful thing to avoid potential re-imports of all the data. We also have another issue where we loose access to gerrit and our apache process pins CPU to 100%, these might also be low hanging friut for people wanting to get involved. [1] http://stackalytics.openstack.org > __ > OpenStack Development Mailing List (not for usage questions) > Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
I wish I'd made it to that Forum session, but here's my two cents now: As a core reviewer for LBaaS I actually find Stackalytics quite helpful for giving me a quick snapshot of contributions, and it lines up almost perfectly in my experience with what I see when I'm actually reviewing and working with people (if you know which statistics to look at -- just sorting by sheer number of reviews or commits and ignoring everything else is of course not useful, and as you say possibly misleading). In all though I actually find that it is a very accurate representation of people's work. For example, in looking at reviewer contributions, I make a mental score based on both the number of reviews, but also the +% (this shouldn't be too high) and the disagreement score (low is generally good, but 0% with a high review count might be questionable). So, I know to discount someone who just spams +1 at everything that has a +2 already and doesn't contribute anything else, which can go unnoticed while reading reviews but sticks out like a sore thumb in Stackalytics. The other side of the coin is someone who posts a ton of useless comments and -1's everything, which then is super obvious to anyone who actually reads reviews. Maybe the experience with the projects I work on is a little different than some of the more populous "base" services like Nova or Neutron? Regardless, I'd be really sad to see it go, as I use it multiple times a week for various reasons. So, I definitely agree with keeping it around and possibly focusing on improving the way the data is displayed. It is definitely best used as one tool in a toolkit, not taken alone as a single source of truth. Is that the main problem people are trying to solve? --Adam On Wed, Jun 7, 2017, 16:38 Ken'ichi Ohmichiwrote: > 2017-05-17 11:55 GMT-07:00 Jeremy Stanley : > > On 2017-05-17 16:16:30 +0200 (+0200), Thierry Carrez wrote: > > [...] > >> we need help with completing the migration to infra. If interested > >> you can reach out to fungi (Infra team PTL) nor mrmartin (who > >> currently helps with the transition work). > > [...] > > > > The main blocker for us right now is addressed by an Infra spec > > (Stackalytics is an unofficial project and it's unclear to us where > > design discussions for it happen): > > > > https://review.openstack.org/434951 > > I also want to find a good design discussion space for stackalytics future. > For example, one of config files is 30KL due to much user information > and that makes the maintenance hard now. > I am trying to separate user part from the existing file but I cannot > find the way to make a consensus for such thing. > In addition, we have two ways for managing bug reports: launchpad and > storyboard if considering it as infra project. > It would be necessary to transport them from launchpad, I guess. > > > In particular, getting the current Stackalytics developers on-board > > with things like this is where we've been failing to make progress > > mainly (I think) because we don't have a clear venue for discussions > > and they're stretched pretty thin with other work. If we can get > > some additional core reviewers for that project (and maybe even talk > > about turning it into an official team or joining them up as a > > deliverable for an existing team) that might help. > > Yeah, diversity will be great for our future. > > Thanks > > __ > OpenStack Development Mailing List (not for usage questions) > Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
2017-05-17 11:55 GMT-07:00 Jeremy Stanley: > On 2017-05-17 16:16:30 +0200 (+0200), Thierry Carrez wrote: > [...] >> we need help with completing the migration to infra. If interested >> you can reach out to fungi (Infra team PTL) nor mrmartin (who >> currently helps with the transition work). > [...] > > The main blocker for us right now is addressed by an Infra spec > (Stackalytics is an unofficial project and it's unclear to us where > design discussions for it happen): > > https://review.openstack.org/434951 I also want to find a good design discussion space for stackalytics future. For example, one of config files is 30KL due to much user information and that makes the maintenance hard now. I am trying to separate user part from the existing file but I cannot find the way to make a consensus for such thing. In addition, we have two ways for managing bug reports: launchpad and storyboard if considering it as infra project. It would be necessary to transport them from launchpad, I guess. > In particular, getting the current Stackalytics developers on-board > with things like this is where we've been failing to make progress > mainly (I think) because we don't have a clear venue for discussions > and they're stretched pretty thin with other work. If we can get > some additional core reviewers for that project (and maybe even talk > about turning it into an official team or joining them up as a > deliverable for an existing team) that might help. Yeah, diversity will be great for our future. Thanks __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [forum] Future of Stackalytics
On 2017-05-17 16:16:30 +0200 (+0200), Thierry Carrez wrote: [...] > we need help with completing the migration to infra. If interested > you can reach out to fungi (Infra team PTL) nor mrmartin (who > currently helps with the transition work). [...] The main blocker for us right now is addressed by an Infra spec (Stackalytics is an unofficial project and it's unclear to us where design discussions for it happen): https://review.openstack.org/434951 In particular, getting the current Stackalytics developers on-board with things like this is where we've been failing to make progress mainly (I think) because we don't have a clear venue for discussions and they're stretched pretty thin with other work. If we can get some additional core reviewers for that project (and maybe even talk about turning it into an official team or joining them up as a deliverable for an existing team) that might help. -- Jeremy Stanley signature.asc Description: Digital signature __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
[openstack-dev] [forum] Future of Stackalytics
Hi! Following the forum, the session moderators should summarize how their session went and detail next steps, to give opportunity to those who couldn't be around in person to catch up if interested. Here is my quick summary of the "Should we kill Stackalytics" session. You can find the etherpad at [1]. After a quick intro on the history of Stackalytics, we spent some time discussing the issues with the current situation. Beyond limited maintenance resources, some accuracy issues and partial infra transition, it became clear in the session that the major driver for a change is that Stackalytics incentivizes the wrong behavior(s), especially in new contributors and organizations in their first step of involvement. As the only "official" and visible way to measure contribution, it encourages dumping useless patches and reviews, and does not value strategic contributions over more tactical contributions. Beyond wasted resources and being annoying to core reviewers, it fails to drive those new contributors to what could be extremely useful contributions, and prevents them to step up in the community. At the same time, several people in the room raised that they would rather not support solutions that would just "kill Stackalytics". Having access to raw metrics on project contribution is useful, for various profiles. It's when you start adding apples to oranges, or deriving rankings from those compound metrics that things start to go very wrong. If Stackalytics was just removed, no doubt it would soon be reborn in other forms elsewhere. Also removing it while not providing anything to replace it is totally useless. We need to first provide a clear incentive to work on desirable items and under-staffed critical teams. The TC is exploring how we could produce such a "help wanted" list (action driven by myself), together with how to give proper recognition to the individuals working on that and/or the organizations funding their work. Once that is done, we'll likely explore how to remove most of the misleading rankings and graphs from Stackalytics to focus it on raw metric information. Before we can do that, we need to complete transition of Stackalytics to OpenStack infrastructure. Once that work is completed it will be time to reconsider our options and have a follow-up discussion session. In summary, the following follow-on work items were identified: 1. Set up the "help wanted" list at TC level and associated hall of fame (action: ttx and TC members) 2. Complete Stackalytics migration to infra (action: infra team, mrmartin) 3. Explore which are the most misleading information/graphs from stackalytics that we might want to remove 4. Reconsider the issue once that work is completed If interested, please jump in! In particular we need help with completing the migration to infra. If interested you can reach out to fungi (Infra team PTL) nor mrmartin (who currently helps with the transition work). [1] https://etherpad.openstack.org/p/BOS-forum-should-we-kill-stackalytics -- Thierry Carrez (ttx) __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev