- Zuul-discuss - lists.opendev.org

Testing multiple commits in one job with Gerrit
by Brendan Jackman 28 Nov '18

28 Nov '18

Hi all, I'm gradually setting up Zuul for my org, and basically I have just one impedance mismatch between Zuul's behaviour and my requirements: Due to pressure on our test hardware I don't think running full end-to-end tests on every individual commit will scale. What I'd ultimately like is a check/gate setup where jobs in the check pipeline runs on every single commit, but jobs in the gate pipeline only run once for an entire "stack" of commits, and then that entire stack can be merged … [View More]at once. IIUC this is basically the behaviour you get already with GitHub, where the zuul notion of "change" maps to a whole Pull Request rather than an individual commit. I started looking into ways something similar can be achieved with Gerrit. After some comments from #zuul and starting at the code a bit this strategy is the one I'm considering: enable one Zuul change to encapsulate multiple Gerrit changes. (Gerrit and Zuul both use the word "change", which works because Gerrit changes and Zuul changes are currently 1:1. That wouldn't be the caes any more with my proposal so I'll always say "Gerrit change" or "Zuul change"). I haven't figured out exactly how the Gerrit changes should be grouped together, and how this should be configured in the Gerrit trigger, but I think the basic idea would be to inject extra events into the Gerrit event stream (somewhere around GerritEventConnector._handleEvent). One example might be that we detect when an entire stack of Gerrit changes has received some common set of approvals, we inject an event called (say) "zuul-change-stack-approved". The user can then have a pipeline.trigger.gerrit entry with this event, and that trigger would create a Zuul change encapsulating the entire stack of Gerrit changes. These wouldn't need to actually be treated in the code like events coming from Gerrit, but this basic idea would allow expressing this behaviour with the existing structure of the Gerrit trigger configuration. I think it would then be a matter of finding the places where the Scheduler and pipeline Managers interact with the Gerrit driver using references to Zuul changes, and ensuring that those parts of the Gerrit driver can handling being passed the new class of Zuul change. I think most of these would involve checking for if the Zuul change is of the new type and then iterating over the constituent Gerrit changes to performe some action against each (e.g. GerritConnection.canMerge, GerritRefFilter.matches, and, with a little more complication, GerritReporter.report). Does that seem like a reasonable approach to the problem? Cheers, Brendan [View Less]

3 6

Multi-repo merge
by Mike Pagel 27 Nov '18

27 Nov '18

Zuul is great in gating projects consisting of many repositories. It even supports dependencies of changes within a single or across multiple repositories. Our project is making heavy use of this. Every now and then we are facing dependencies that unfortunately are not unidirectional. I know, we should have better thought about clean repo structures in the first place, but one not too artificial example is a situation where repositories are structured to express ownership, visibility and write … [View More]

6 12

Zuul CI and Gitlab
by Алексей Степаненко 26 Nov '18

26 Nov '18

Hi. At the summit in Berlin was interested in how the Zuul CI system. GitHub support has been announced, but we're using GitLab. Will there be added support for Networks in the future ? -- Aleksei Stepanenko, Selectel Team

1 0

Semaphores and resources
by Tobias.Henkel＠bmw.de 26 Nov '18

26 Nov '18

Hi Zuulers, We just discovered an (in our case) unpleasant behaviour of zuul and semaphores. Scenario: We have a job ‘foo’ with a semaphore of 5. Now the queue grows and we have 30 items with job ‘foo’ in the queue. The current behaviour of zuul is that it first gets the nodes and then locks the semaphore. This means that 5 jobs run and 25 wait for the semaphore but already got and locked their nodes. In case the job is expensive and long running this can become quite a big resource waste … [View More]

5 7

Dequeuing buildset on first error
by Tobias.Henkel＠bmw.de 19 Nov '18

19 Nov '18

Hi Zuulers, We have several projects that have buildsets with fast (few minutes, e.g. linters) and longer running (> 40 minutes, e.g. big fat c++ compilation) jobs. We often see the case that one of the linter jobs failed and already could give feedback. However zuul waits for the completion of the fat compile job before reporting. Generally this is the desired behaviour because we want to give as much feedback as possible. However some projects may want to decide that quick feedback is … [View More]

7 6

Slides from Tungsten talk on Berlin Summit
by Jarek Łukow 15 Nov '18

15 Nov '18

Hi all, As promised, I'm sending the link to our slide deck: https://wiki.tungsten.io/download/attachments/1409104/Zuul_as_a_build_syste… Regards, -- Jarosław Łukow DevOps Engineer jaroslaw.lukow(a)codilime.com -- ------------------------------- The information in this email is confidential and may be legally privileged, it may contain information that is confidential in CodiLime Sp. z o. o. It is intended solely for the addressee. Any access to this email by third parties is … [View More]

2 1

OpenDev, the future of OpenStack Infra
by Clark Boylan 08 Nov '18

08 Nov '18

Hello everyone, Sorry for another cross post so soon. In the land before time we had Stackforge. Stackforge gave non-OpenStack projects a place to live with their own clearly defined "not OpenStack" namespacing. As the wheel of time spun we realized that many Stackforge projects were becoming OpenStack projects and we would have to migrate them. This involved Gerrit downtimes to rename things safely. To ease the pain of this, the TC decided that all projects developed in the OpenStack … [View More]Infrastructure could live under the OpenStack git namespace to simplify migrations. Unfortunately this had the effect of creating confusion over which projects were officially a part of OpenStack, and whether or not projects that were not OpenStack could use our project hosting. Stackforge lived on under a different name, "unofficial project hosting", but many potential infrastructure users either didn't understand this or didn't want that strong association to OpenStack for their project hosting [0]. Turns out that we want to be able to host OpenStack and non-OpenStack projects together without confusion in a way that makes all of the projects involved happy. In an effort to make this a reality the OpenStack Infra team has been working through a process to rename itself to make it clear that our awesome project infrastructure and open collaboration tooling is community run, not just for OpenStack, but for others that want to be involved. To this end we've acquired the opendev.org domain which will allow us to host services under a neutral name as the OpenDev Infrastructure team. The OpenStack community will continue to be the largest and a primary user for the OpenDev Infrastructure team, but our hope in making our infrastructure services more inclusive is that we'll also attract new contributors, which will ultimately benefit OpenStack and other open infrastructure projects. Our goals for OpenDev are to: * Encourage additional active infrastructure contributors to help us scale. Make it clear that this is community-run tooling & infrastructure and everyone can get involved. * Make open source collaboration tools and project infrastructure more accessible to those that want it. * Have exposure to and dogfooding of OpenStack clouds as viable open source cloud providers. * Enable more projects to take advantage of the OpenStack-pioneered model of development and collaboration, including recommended practices like code review and gating. * Help build relationships with new and adjacent open source projects and create an inclusive space for collaboration and open source development. Much of this is still in the early planning stages. This is the infrastructure team's current thinking on the subject, but understand we have an existing community from which we'd like to see buy-in and involvement. To that end we have tried to compile a list of expected FAQ/Q&A information below, but feel free to followup either on this thread or with myself for anything we haven't considered already. Any transition will be slow and considered so don't expect everything to change overnight. But don't be surprised if you run into some new HTTP redirects as we reorganize the names under which services run. We'll also be sure to keep you informed on any major (and probably minor) transition steps so that they won't surprise you. Thank you, Clark [0] It should be noted that some projects did not mind this and hosted with OpenStack Infra anyway. ARA is an excellent example of this. FAQ * What is OpenDev? OpenDev is community-run tools and infrastructure services for collaboratively developing open source software. The OpenDev infrastruture team is the community of people who operate the infrastructure under the umbrella of the OpenStack Foundation. * What services are you offering? What is the expected timeline? In the near-term we expect to transition simple services like etherpad hosting to the OpenDev domain. It wil take us months and potentially up to a year to transition key infrastructure pieces like Git and Gerrit. Example services managed by the team today include etherpad, wiki, the zuul and nodepool CI system, git and gerrit, and other minor systems like pbx conferencing and survey tools. * Where will these services live? We've acquired opendev.org and are planning to set up DNS hosting very soon. We will post a simple information page and FAQ on the website and build it out as necessary over time. * Why are you changing from the OpenStack infrastructure team to the OpenDev infrastructure team? In the same way we want to signal that our services are not strictly for OpenStack projects, and that not every project using our services is an official part of OpenStack, we want to make it clear that our team also serves this larger community. * Who should use OpenDev services? Does it have to be projects related to OpenStack, or any open source projects? In short, open source contributors who share our community values, especially those who might want to help contribute to improving and maintaining OpenDev infrastructure over time. Projects using OpenDev hosted git and gerrit services should have an OSI-approved license. * Will the OpenStack projects live at git.opendev.org? All projects hosted with OpenDev will live at git.opendev.org. For backwards compatibility reasons, at the very least git.openstack.org will be an alias for git.opendev.org for the forseeable future. The same is true of the other existing whitelabel git domains such as git.starlingx.io and git.zuul-ci.org. Whether or not other 'whitelabel' domains are created is an open question. Given a neutral domain name, the desire for such sites may not seem as necessary. * Does this mean the infrastruture team will be spending less time on OpenStack? OpenStack will continue to be the largest and a primary user for the OpenDev Infrastructure team, and we expect that our work will benefit all users. There will be additional effort required as we transition to the new namespace and reorganize, but over the long term we hope this inclusive approach will help us attract new contributors and ultimately benefit OpenStack. * Are OpenStack cloud test resources the only resources that will be used? At the present time all of the donated resources come to us from a combination of OpenStack Public and Private clouds. Nobody from any of the proprietary clouds has asked to donate resources to us. It is conceivable that the shift to OpenDev could open the door to those cloud providers wanting to donate some cloud resources. Assuming nodepool supports talking to those clouds, it is certainly a possibility, but at the moment it's all speculation. * Is this name associated with the OpenDev Conferences (opendevconf.com) that OpenStack Foundation has previously organized? Yes! They are related, albeit indirectly. To quote from the conference's promotional site, "the focus is on bringing together composable open infrastructure technologies across communities and industries." The possibility of cross-promotional tie-ins could prove synergistic, since the conference and the collaboratory we're building share a lot of similar values and goals, and are ultimately supported by the same donors, community and foundation. * How will OpenDev be governed? Will the OpenStack TC retain oversight over it? The OpenDev governance discussion is just getting started, but like all OSF-supported initiatives, OpenDev follows the Four Opens, so it will ultimately be directly governed by OpenDev contributors. While it won't be under the sole oversight of the OpenStack Technical Committee anymore, OpenDev users (in particular OpenStack) should be represented in the governance model so that they can feed back their requirements to the OpenDev team. [View Less]

1 0

Community Infrastructure Berlin Summit Onboarding Session
by Clark Boylan 05 Nov '18

05 Nov '18

Hello everyone, My apologies for cross posting but wanted to make sure the various developer groups saw this. Rather than use the Infrastructure Onboarding session in Berlin [0] for infrastructure sysadmin/developer onboarding, I thought we could use the time for user onboarding. We've got quite a few new groups interacting with us recently, and it would probably be useful to have a session on what we do, how people can take advantage of this, and so on. I've been brainstorming ideas on this … [View More]

1 0

Node request priorities
by corvus＠inaugust.com 31 Oct '18

31 Oct '18

Hi, In OpenStack, we occasionally have periods where, for various reasons, one project, or a group of projects, uses a sufficiently large number of node resources that it significantly delays results for other projects. We happen to be in one of those periods, so I've had some time to think about ways to mitigate this. First, let me say that one of the potential causes is long and unreliable tests. There are efforts underway to communicate that and hopefully address it. But that topic isn't … [View More]really for this list. Our concern is, when a system gets into this state, can or should we do anything in Zuul to mitigate it? I would like to avoid (for now) the use of project-based quotas. I could certainly see us adopting tenant quotas in the future, but that doesn't help us in this case where all of the contention is within a single tenant. Within a tenant, project quotas just seem to be too much work. I wouldn't even begin to know how to set them fairly. And what we really want to do is use as many node resources are available, but use them in a fair manner. I think we can accomplish this by augmenting our use of node request priority. Currently, we use a priority queue with very coarse granularity (high, normal, low) set by Zuul according to pipeline. I'd like to adjust the system so that each subsequent node request from a given project has a lower priority. In other words, at any given time, the first change from project A in gate has the highest priority, and the same is true for project B. The second change from A has a slightly lower priority, and so forth. The problem with this is that the node request priorities much change over time as the contents of the pipeline queues change. Once the first change from A merges, the second change from A should now have the highest priority. That makes this incompatible with our current priority queue system. Fortunately, Tobias has been working on adding caching to Nodepool's use of Zookeeper. If we start having node requests cached in-memory in the launchers, the in-memory records can be automatically updated by Zookeeper callbacks. This means that we can encode our additional priority information on an attribute in the Znode, updating it as necessary, and combine it with the already existing information from the priority queue to determine the order in which to fulfill requests. To summarize, I propose the following: * The current priority queue system for node requests will remain in place. I'll call that "precedence" to distinguish it from the new bits. * Add a "priority" field to the node request znodes in Zookeeper. This is an integer which indicates the priority of this request relative to others at the same pipeline precedence level. * Zuul will update the priority field after each time a pipeline queue processor runs, according to internal heuristics. I propose that initially we decrement the priority for each subsequent change from the same project, though we may also add an option to do so by project-group, as determined by membership in a shared change queue. * We add temporary feature flags to Zuul so that we can try out these new features (including different options for how to determine the priority) before committing to future support. Whatever we decide to keep, we may also want to make optional, so these flags may eventually turn into permanent configuration flags. * We implement a TreeCache for node requests in nodepool. * When nodepool begins to iterate over all node requests in order to handle them, it first sorts by precedence, then priority. Items with the same precedence and priority will remain in their chronological order of submission. I think that will help make node distribution more fair, and it builds on top of the current system without changing it so it should be forward/backward compatible, and even optional. -Jim [View Less]

5 8

Discussion between OSF projects in Berlin
by Thierry Carrez 31 Oct '18

31 Oct '18

Hi! For those who will be in Berlin for the Summit, please mark your calendars for the following Forum session: https://www.openstack.org/summit/berlin-2018/summit-schedule/events/22815/c… This session is an opportunity for the technical leadership of all OSF projects to meet, discuss pain points, compare experiences and explore synergies. We expect to have members of the OpenStack, Kata Containers, StarlingX, Airship, and Zuul communities around. It will be the occasion, among other … [View More]

1 0