Rally/Updates
Contents
Weekly updates
September 26, 2014
Hello stackers,
here is a brief overview of what has happened in Rally recently:
- We have refactored the code responsible for atomic actions processing in Rally benchmarks. Let's remind you that each benchmark scenario in Rally consists of a series of atomic actions, whereas the running time of each atomic action is measured in the same way as that of the whole scenario. After refactoring, we now ensure that Rally doesn't skip atomic actions that failed and distinguishes different runs of two atomic actions with the same name (see an example of how a typical results table could look before and looks after refactoring).
- Another direction of refactoring was the SLA code; it has been modified so that now SLA results are stored in the DB along with other benchmarking data.
- Check out a nice user story about VMs boot performance in Nova with Rally. It shows how Rally can be used in practice to catch reals bugs and improve OpenStack performance.
- We work hard on achieving 100% test coverage in Rally: last week, additional unit tests for contexts and scenarios have been merged.
Our current priorities are further refactoring steps, including those in critical parts of Rally (e.g. temporary user creation and cloud cleanup code); we also strive towards making Rally bug-free and continuously issue different bugfixing patches.
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
September 15, 2014
Hello stackers,
there has been much diverse and useful contribution to Rally recently. Let us highlight some interesting updates in our project:
- Rally is on its way to support of benchmarking OpenStack clouds using ordinary user accounts that already exist. Rally lacked such functionality (it only supported benchmarking either from an admin account or from a bunch of temporarily created users), which posed a problem since some deployments don't allow temporary users creation. There have been two patches that prepare the code for this new functionality. It is going to come very soon - stay tuned.
- We constantly improve our gate jobs (jobs that run a test suite for every patch pending review in Rally). Recently, we have introduced a nice aggregated results page for that gate job (see an example). It now makes it very easy for developers to navigate through the results of tests agains their patches to Rally.
- We have added a new "volumes" context which makes it possible to create cinder volumes in the benchmark environment and use them later in the actual benchmark scenarios.
- Much work has been accomplished on the overall code quality improvement. We have both made several code refactoring patches and introduced several new test suites (say, functional tests for CLI).
Current work centers around introducing benchmarking with pre-created users as mentioned above; much work is currently devoted to new benchmark scenarios as well. Also please note that there will be soon an important update on Rally ReadTheDocs page, which will make it much easier to navigate and get through, especially for newbies.
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
September 1, 2014
Hello stackers,
over the past week, Rally has been extended with the following features:
- Rally now offers a pretty simple feature request mechanism. Everyone who is interested in adding new functionality to Rally now can share his ideas in a standartized format (see an example). All you need to do is to write down your feature request in a separate rst-file in the doc/feature_request folder and submit it as a patch to Rally (if you are unsure about how to do this, read our "How to contribute" tutorial).
- We added a new benchmark scenario for Cinder quotas creation/deletion.
- Designate support in Rally has been extended so that it is posisble now to set up Designate quotas.
There is pretty much work going on this week, including the introduction of benchmarking with existing users (in addition to those created temporarily, which was the only option before), and also improvements in the plugin mechanism, so that it will be possible now to write plugins for runners and contexts. We do lots of code refactoring and will do even more in the upcoming weeks.
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
August 25, 2014
Hello stackers,
here is a quick overview of what has been added to Rally recently:
- Rally now supports Designate, which is a DNaaS service for OpenStack (providing REST API for domain/record management). Several new bechmark scenarios for Designate have been added as well;
- In the CLI, we have added a new command called rally info. It is a essentially a special search engine embedded into Rally, which, for a given search query, prints documentation for the corresponding benchmark scenario/deploy engine/... as fetched from the source code. Thus you can learn about different Rally entities without leaving the Command Line Interface. For usage samples, see this link;
- We have extended the SLA output which indicates whether a benchmark has passed some set of predefined success criteria with customizable messages;
- We have performed a range of nice optimizations:
The main work directions for the next week comprise further improvement in the contexts mechanism (both the algorithm of context creation and new context classes), further optimizations (e.g., multithreaded objects deletion) and also code refactoring (e.g., of that for tracking atomic actions in benchmark scenarios)
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
August 4, 2014
Hello stackers,
below you will find the most recent updates in Rally:
- The periodic scenario runner has been refactored and renamed, as of now, to rpc (runs per second). Note that this renaming also affects task configuration files that use that runner type. The runner itself now has been reimplemented with the multiprocessing.Process class (instead of multiprocessing.Pool), which potentially decreases memory usage on large iterations and reduces errors.
- A nice optimization to the chart generation code ensures that the Rally plots will be rendered fast in browsers even for tasks with a huge number of iterations completed.
- New benchmark scenarios include those testing the Nova server resize operation and a set of benchmarks for Sahara group node templates.
- We continue extending our gates with nice features. A new one are the SLA checks that add information about whether benchmark scenarios pass a set of success criteria (see an example)
Many interesting improvements are on their path to being merged to Rally. Among them, let us mention the "rally info" command, which prints descriptions for different entities in Rally to the console, further work on the Rally documentation which gets cleaner and cleaner, and also a set of new context classes, including those for generating images for VM benchmarks, as well as for creating Neutron networks.
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
July 17, 2014
Hello stackers,
here are some recent updates in Rally:
- Results of scenario runners are now stored asynchronously. This is precondition for such further features like piecemeal storing of results to DB (to reduce RAM usage), or progress displaying.
- Add check if required services are available before starting the scenario. This is realized by adding services validation to scenarios. Also this patch includes new validation mechanism.
- Ability to cancel task by keyboard interrupts is availiable now;
- Add base to provide interation testing;
- A lot of of patches to improve unit tests coverage.
This week, our work continues with such main novelties as adding support for Keystone API v3, adding context classes for avoiding vm creation if tenant has no network and prepare an image that will have installed the required programs, mechanism to provide rally as a service, code refactoring in context manager to make code more clear.
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
July 07, 2014
Hello stackers,
here is the update for the last weeks. From all the work we've completed we would like to highlight the following:
- New benchmark scenarios:
- Benchmark to validate a keystone token N times at service endpoint that allows to check effect of caching related to tokens;
- Scenario for benchmark Neutron ports.
- Add nova floating ip management in VM scenario, before this patch the VM runcommand scenario used the fixed ip to ssh connect to the instance, this could only worked if fixed ip range was directly accessible with was the case in very limited deployments.
- Add image context class that allows adding images to each user for benchmarks, also this patch provides scenario which tests glance image-list command;
- Large update of documentation that simplifies its index structure and headings, extends the 'Benchmark' page with info from the "Main concepts" and adds some introductory text on Deploy & Verify parts;
- Add service-level agreement checking that allows to add section in task configuration that contains the criteria of success, e.g "less then 5% of failure rate", or "faster then 5 seconds for one iteration";
- Use oslo.db module and remove rally/openstack/common/db.
Current work includes such interesting novelties as context semantic validation, periodic runner refactoring, "rally info" command which prints descriptions for different entities in Rally, checking if required services are available before starting the scenario and refactoring of validation system, continue work on "stress" runner.
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
June 23, 2014
Hello stackers,
here is a quick overview of the major contribution in Rally made over the past week:
- New benchmark scenarios:
- Neutron routers API;
- Ceilometer scenario fixes (e.g. the CeilometerBasic scenario group has now been renamed to CeilometerAlarms).
- A quite nice work on code cleanup according to the new hacking 0.9.x requirements has been done to be able to update Rally's global requirements;
- Rally gates have been prepared to be able to run VM scenarios (this work will be finished in the near future).
Current work includes such interesting novelties as the new "stress" scenario runner that generates a growing load until it breaks the cloud, new contexts: Image (boots VMs before benchmarking) and NeutronNetwork (adds pre-created networks) and many others. Also note that there will be soon a large update in the ReadTheDocs Rally page that will make it much more user-oriented.
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
June 16, 2014
Hello stackers,
during the past week, our work has primarily been concentrated on minor bugfixes and enhancements in different parts of Rally. These include:
- Aggregated tables for atomic action runtimes are now available not only in the CLI, but also in the HTML output (generated by the $ rally task plot2html ... command). Let us remind that atomic actions are basic procedures that benchmark scenarios consist of (such as booting a server in Nova or deleting a user in Keystone).
- The docs section of Rally has been extended with the so-called "User stories", which describe use cases of Rally in the real world, some results obtained by Rally as well as their analysis. The first user story is about token creation under different load in Keystone.
- Bugfixes include:
- Fixed installation failure in case pip is not installed on the target system;
- Bugfixes in the quotas context introduce namings corrections and also catching possible errors in the cleanup procedure;
- The constant_for_duration scenario runner now stores results for all the benchmark iterations it has performed, not only for the last ones.
Current work includes such interesting novelties as the new "stress" scenario runner that generated a growing load until it breaks the cloud, the new Image context that boots VMs before benchmarking and many others.
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
June 2, 2014
Hello stackers,
one of the most pleasant news for the past week is the appearance of a brand new Rally ReadTheDocs page, where all the docs from the Rally repository are available in a human-readable format.
As for recent patches merged to Rally, so let's pick up the most interesting ones:
- There is now the new Deploy Engine for FUEL (that, as other deploy engines, deploys an OpenStack installation using FUEL and returns endpoints to it).
- A nice refactoring patch has rewritten much CLI code so that it uses more Oslo code.
- Following our good tradition to extend the set of benchmark scenarios available in Rally each week, there are now scenarios for Ceilometer Statistics API.
Among new interesting patches being on their way to Rally, let's mention a filesystem benchmark (that will be launched inside created servers), new context classes for creating images and Neutron networks before launching becnhmarks, and many others.
We encourage you to take a look at new patches in Rally pending for review and to help us make Rally better.
Source code for Rally is hosted at GitHub: https://github.com/stackforge/rally
You can track the overall progress in Rally via Stackalytics: http://stackalytics.com/?release=juno&metric=commits&project_type=all&module=rally
Open reviews for Rally: https://review.openstack.org/#/q/status:open+rally,n,z
Stay tuned.
Regards,
The Rally team
