John Griffith: Project Technical lead
IRC: jgriffith
LaunchPad: john-griffith
State of OpenStack Block Storage
A look at what’s planned for Juno (Jun 24, 2014)
Theme for Juno:
Some of the feedback we’ve received from
“Enterprise” users:
● Continue to improve backups
● Focus on compatibility
● Don’t break upgrades
● Keep giving me something that works
● I need to be able to use my existing gear
● Quality is key
● Need an HA story for the reference driver
Improve Backups:
● We currently have a backup to SWIFT (and some others)
● Problem is it’s a full copy of the Volume only
● Difficult to Restore
● More difficult to import (ie DR)
➔Working on adding incrementals for Juno
➔Improving restore including “mange/unmanage”
Compatibility:
This means different things to different people
● Compatibility of Cinder versions between upgrades
● Compatibility of features between different backend devices
● MOST importantly, continued compatibility of the reference LVM
driver
➔Goal is for all 3’rd party drivers to implement CI (dsvm-full)
➔Ensures core API functions are tested/supported on all drivers
➔Public visibility of testing and compatibility
➔DONT BREAK THINGS BETWEEN VERSIONS
Compatibility Plans for Juno:
● Goal is for all 3’rd party drivers to implement CI (dsvm-full)
● Ensures core API functions are tested/supported on all drivers
● Public visibility of testing and compatibility
● Don’t make it hard for me to upgrade
● DONT BREAK THINGS BETWEEN VERSIONS
Something that works:
● See previous slide
● Functions and capabilities should be expected regardless of
backend
● Reference LVM implementation is still the primary focus
Using my existing gear:
Common theme seems to be “repurposing gear for OpenStack”
● Vendor participation is good if it’s compatible and it works
● I like having choices
● I like mixing and matching
● I might change my mind in the future
● DONT LOCK ME IN
HA for reference implementation:
Multiple paths we hope to work on in Juno
● DRBD based solution
● Volume mirroring across multiple cinder-volume nodes using
software raid
What we’re NOT hearing:
● None of the features I want are here
● There are too many options
● I don’t need Block Storage
● You’re missing ‘xyz’
Some other things we’re working on:
● Significant code cleanup and fixes in “object sprawl”
● Ability to help admins set up replication
● Concept of consistency groups between volumes
● Intelligently understand devices that use the concept of storage
pools
● Scheduling local storage on Compute Nodes*
What are we missing:
If you’re deploying a Private Cloud with OpenStack (or considering
it)
● What do you need from Cinder that’s missing?
● What problems do you have with Cinder currently?
● How can we help make it easier, and more beneficial for you to
use?
DON’T HESITATE TO REACH OUT TO ME!!!
Get Involved:
Contributing to OpenStack isn’t just for developers
Open Source is more than just writing code
● Documentation
● Bug reports
● Feature requests
● Feedback, Feedback, Feedback
● Supporting others
Questions?
email: john.griffith@solidfire.com
twitter: @jdg_8
IRC: jgriffith

Block Storage Updates - Juno Edition

  • 2.
    John Griffith: ProjectTechnical lead IRC: jgriffith LaunchPad: john-griffith State of OpenStack Block Storage A look at what’s planned for Juno (Jun 24, 2014)
  • 3.
  • 4.
    Some of thefeedback we’ve received from “Enterprise” users: ● Continue to improve backups ● Focus on compatibility ● Don’t break upgrades ● Keep giving me something that works ● I need to be able to use my existing gear ● Quality is key ● Need an HA story for the reference driver
  • 5.
    Improve Backups: ● Wecurrently have a backup to SWIFT (and some others) ● Problem is it’s a full copy of the Volume only ● Difficult to Restore ● More difficult to import (ie DR) ➔Working on adding incrementals for Juno ➔Improving restore including “mange/unmanage”
  • 6.
    Compatibility: This means differentthings to different people ● Compatibility of Cinder versions between upgrades ● Compatibility of features between different backend devices ● MOST importantly, continued compatibility of the reference LVM driver ➔Goal is for all 3’rd party drivers to implement CI (dsvm-full) ➔Ensures core API functions are tested/supported on all drivers ➔Public visibility of testing and compatibility ➔DONT BREAK THINGS BETWEEN VERSIONS
  • 7.
    Compatibility Plans forJuno: ● Goal is for all 3’rd party drivers to implement CI (dsvm-full) ● Ensures core API functions are tested/supported on all drivers ● Public visibility of testing and compatibility ● Don’t make it hard for me to upgrade ● DONT BREAK THINGS BETWEEN VERSIONS
  • 8.
    Something that works: ●See previous slide ● Functions and capabilities should be expected regardless of backend ● Reference LVM implementation is still the primary focus
  • 9.
    Using my existinggear: Common theme seems to be “repurposing gear for OpenStack” ● Vendor participation is good if it’s compatible and it works ● I like having choices ● I like mixing and matching ● I might change my mind in the future ● DONT LOCK ME IN
  • 10.
    HA for referenceimplementation: Multiple paths we hope to work on in Juno ● DRBD based solution ● Volume mirroring across multiple cinder-volume nodes using software raid
  • 11.
    What we’re NOThearing: ● None of the features I want are here ● There are too many options ● I don’t need Block Storage ● You’re missing ‘xyz’
  • 12.
    Some other thingswe’re working on: ● Significant code cleanup and fixes in “object sprawl” ● Ability to help admins set up replication ● Concept of consistency groups between volumes ● Intelligently understand devices that use the concept of storage pools ● Scheduling local storage on Compute Nodes*
  • 13.
    What are wemissing: If you’re deploying a Private Cloud with OpenStack (or considering it) ● What do you need from Cinder that’s missing? ● What problems do you have with Cinder currently? ● How can we help make it easier, and more beneficial for you to use? DON’T HESITATE TO REACH OUT TO ME!!!
  • 14.
    Get Involved: Contributing toOpenStack isn’t just for developers Open Source is more than just writing code ● Documentation ● Bug reports ● Feature requests ● Feedback, Feedback, Feedback ● Supporting others
  • 15.