Meetings/Storlets

Meeting Time: Every Wednesday at 13:00 UTC in #openstack-storlets

Agenda:

• Prioritize patches (suggestion):
  • No patches now ready for review
  • Incoming patches
    • Use contextmanager to manager file descriptors: https://review.openstack.org/#/c/282211/
      • COPY should be fixed
    • Refactor directory structure about python modules: https://review.openstack.org/#/c/285947/
      • Needs more work about fixing setup.py and rebasing
      • We should decide when we rebase/merge this, to avoid many merge conflicts caused by it. [takashi]
      • Move swift specific functionality from gateway to middleware

• Storlet in Hackathon etherpad
  • I would like to add another analytics related question.
  • What timeframe do we think we need?

Austin:

• We currently have the following abstract
  • Title: From Analytic to Image Processing in Swift with Openstack Storlets
  • Openstack Stolets is an emerging Openstack project aimed at pushing down data centric compute jobs to the object store. Storlets offer a server-less programming model enabling developers to quickly develop and deploy data path computations without worrying about the underlying platform details.
  • Storlets facilitate several interesting use cases such as:
    • Performing image processing in the storage, allowing to process large objects without downloading it.
    • Spark analytic push down to the store, saving on both bandwidth and memory in the analytic cluster.
    • Data obfuscation and anonymization at the source. Thus, preventing sensitive data from ever leaving the store.
  • In this talk we will present the project, discuss and demonstrate the use cases and present a PoC about an effective architecture for image processing using Storlets done at NTT.

Swift Hackathon

• See the following etherpad
  • https://etherpad.openstack.org/p/swift-hackathon-feb-2016-storlets

Bugs

• Range handling: fix released
• FD leak
  • FD leak in Storlet GET: https://bugs.launchpad.net/storlets/+bug/1537982 (Fix Released)
    • https://review.openstack.org/#/c/272379/ ... merged
  • There are other possible FD leaks, which may happen in timeout/exception case.
    • Use context manager to make sure all FDs are closed: https://review.openstack.org/#/c/282211/ (patch should be fixed)
• Validate metadata in POST/COPY about starlet object: https://bugs.launchpad.net/storlets/+bug/1555411
  • POST is already fixed: https://review.openstack.org/#/c/290952/
  • How can we validate metadata when COPYing object outside starlet container into starlet container?