yahoo-eng-team team mailing list archive

[Brian Haley <1274536@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

We've noticed that under high loads - 100's of VMs booting
simultaneously from multiple tenants, that the metadata agent simply
can't keep up, and the VMs fail to get their metadata.

Debug showed that the listen backlog was being overflowed, but changing
that manually only made things a little better.  It wasn't until
multiple threads were started that the problem went away.

Opening this bug to get these upstream since they help with the
scalability of Neutron.

1) Change the metadata-agent to support having multiple worker threads
2) Change it to support a configurable listen backlog

I'll assign this to myself and get the changes out for review.

** Affects: neutron
     Importance: Undecided
     Assignee: Brian Haley (brian-haley)
         Status: New

** Changed in: neutron
     Assignee: (unassigned) => Brian Haley (brian-haley)

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1274536

Title:
  Neutron metadata agent can't handle high loads

Status in OpenStack Neutron (virtual network service):
  New

Bug description:
  We've noticed that under high loads - 100's of VMs booting
  simultaneously from multiple tenants, that the metadata agent simply
  can't keep up, and the VMs fail to get their metadata.

  Debug showed that the listen backlog was being overflowed, but
  changing that manually only made things a little better.  It wasn't
  until multiple threads were started that the problem went away.

  Opening this bug to get these upstream since they help with the
  scalability of Neutron.

  1) Change the metadata-agent to support having multiple worker threads
  2) Change it to support a configurable listen backlog

  I'll assign this to myself and get the changes out for review.

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1274536/+subscriptions