rally/doc/release_notes/v0.0.2.rst

Rally v0.0.2

Information

Commits	100
Bug fixes	18
Dev cycle	45 days
Release date	12/Mar/2015

Details

This release contains new features, new benchmark plugins, bug fixes, various code and API improvements.

New Features

• rally task start --abort-on-sla-failure

  Stopping load before things go wrong. Load generation will be interrupted if SLA criteria stop passing.

• Rally verify command supports multiple Tempest sources now.

• python34 support

• postgres DB backend support

API changes

• [new] rally [deployment | verify | task] use subcommand

  It should be used instead of root command rally use

• [new] Rally as a Lib API

  To avoid code duplication between Rally as CLI tool and Rally as a Service we decide to make Rally as a Lib as a common part between these 2 modes.

  Rally as a Service will be a daemon that just maps HTTP request to Rally as a Lib API.

• [deprecated] rally use CLI command

• [deprecated] Old Rally as a Lib API

  Old Rally API was quite mixed up so we decide to deprecate it

Plugins

• Benchmark Scenario Runners:

  [improved] Improved algorithm of generation load in constant runner

  Before we used processes to generate load, now it creates pool of processes (amount of processes is equal to CPU count) after that in each process use threads to generate load. So now you can easily generate load of 1k concurrent scenarios.

  [improved] Unify code of constant and rps runners

  [interface] Added abort() to runner's plugin interface

  New method abort() is used to immediately interrupt execution.

• Benchmark Scenarios:

  [new] DesignateBasic.create_and_delete_server

  [new] DesignateBasic.create_and_list_servers

  [new] DesignateBasic.list_servers

  [new] MistralWorkbooks.list_workbooks

  [new] MistralWorkbooks.create_workbook

  [new] Quotas.neutron_update

  [new] HeatStacks.create_update_delete_stack

  [new] HeatStacks.list_stacks_and_resources

  [new] HeatStacks.create_suspend_resume_delete_stac

  [new] HeatStacks.create_check_delete_stack

  [new] NeutronNetworks.create_and_delete_routers

  [new] NovaKeypair.create_and_delete_keypair

  [new] NovaKeypair.create_and_list_keypairs

  [new] NovaKeypair.boot_and_delete_server_with_keypair

  [new] NovaServers.boot_server_from_volume_and_live_migrate

  [new] NovaServers.boot_server_attach_created_volume_and_live_migrate

  [new] CinderVolumes.create_and_upload_volume_to_image

  [fix] CinderVolumes.create_and_attach_volume

  Pass optional **kwargs only to create server command

  [fix] GlanceImages.create_image_and_boot_instances

  Pass optional **kwargs only to create server command

  [fix] TempestScenario.* removed stress cleanup.

  Major issue is that tempest stress cleanup cleans whole OpenStack. This is very dangerous, so it's better to remove it and leave some extra resources.

  [improved] NovaSecGroup.boot_and_delete_server_with_secgroups

  Add optional **kwargs that are passed to boot server comment

• Benchmark Context:

  [new] stacks

  Generates passed amount of heat stacks for all tenants.

  [new] custom_image

  Prepares images for benchmarks in VMs.

  To Support generating workloads in VMs by existing tools like: IPerf, Blogbench, HPCC and others we have to have prepared images, with already installed and configured tools.

  Rally team decide to generate such images on fly from passed to avoid requirements of having big repository with a lot of images.

  This context is abstract context that allows to automate next steps:

  1. runs VM with passed image (with floating ip and other stuff)
  2. execute abstract method that has access to VM
  3. snapshot this image

  In future we are going to use this as a base for making context that prepares images.

  [improved] allow_ssh

  Automatically disable it if security group are disabled in neutron.

  [improved] keypair

  Key pairs are stored in "users" space it means that accessing keypair from scenario is simpler now:

  self.context["user"]["keypair"]["private"]

  [fix] users

  Pass proper EndpointType for newly created users

  [fix] sahara_edp

  The Job Binaries data should be treated as a binary content

• Benchmark SLA:

  [interface] SLA calculations is done in additive way now

  Resolves scale issues, because now we don't need to have whole array of iterations in memory to process SLA.

  This is required to implement --abort-on-sla-failure feature

  [all] SLA plugins were rewritten to implement new interface

Bug fixes

18 bugs were fixed, the most critical are:

• Fix rally task detailed --iterations-data

  It didn't work in case of missing atomic actions. Such situation can occur if scenario method raises exceptions

• Add user-friendly message if the task cannot be deleted

  In case of trying to delete task that is not in "finished" status users get traces instead of user-friendly message try to run it with --force key.

• Network context cleanups networks properly now

Documentation

• Image sizes are fixed
• New tutorial in "Step by Step" relate to --abort-on-sla-failure
• Various fixes