Bug 1081001 - Need to allow disabling of bandwidth polling periodic task
Summary: Need to allow disabling of bandwidth polling periodic task
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Red Hat OpenStack
Classification: Red Hat
Component: openstack-nova
Version: 4.0
Hardware: Unspecified
OS: Unspecified
medium
low
Target Milestone: z4
: 4.0
Assignee: Vladan Popovic
QA Contact: Vladan Popovic
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2014-03-26 13:21 UTC by Russell Bryant
Modified: 2022-07-09 06:17 UTC (History)
5 users (show)

Fixed In Version: openstack-nova-2013.2.3-4.el6ost
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2014-05-29 20:35:36 UTC
Target Upstream Version:
Embargoed:

Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
OpenStack gerrit 60641 0 None None None Never
Red Hat Product Errata RHSA-2014:0578 0 normal SHIPPED_LIVE Moderate: openstack-nova security, bug fix, and enhancement update 2014-05-30 00:27:05 UTC

Description Russell Bryant 2014-03-26 13:21:47 UTC 
In RHOS 5.0, the periodic task related to the bandwidth_poll_interval configuration option is automatically turned into a no-op when the underlying driver does not support it.  In RHOS 4.0, this task will do unnecessary work every time it runs.  This includes querying the details of all running instances on the compute host from the conductor service.  This adds unnecessary load on the database, conductor, message bus, and the compute host.

This task is only relevant for the xenapi driver so we should backport the patch that turns off the task automatically.


commit 4f82543ac7427638fec7e286bbb84fd7b3e3e9f3
Author: Phil Day <philip.day>
Date:   Fri Dec 6 23:54:43 2013 +0000

    Make it possible to disable polling for bandwidth usage
    
    Bandwidth usage is only supported by some hypervisor drivers, but
    the period task always runs and asks conductor for a list of
    instances before it gets a NotImplementedError from the virt driver.
    
    This change allows this to be disabled by setting bandwidth_poll_interval
    to 0 which is consistent with other periodic task interval settings,
    avoiding the wasted conductor call.
    
    It also will automatically disable bandwidth polling if the driver raises
    NotImplemented.
    
    Change-Id: I2ac9c967c5ceafffc39a0a372146c762891a08b8
Comment 4 Vladan Popovic 2014-05-28 15:36:21 UTC 
Without the patch (by setting the periodic task to 1min):
2014-05-28 11:30:39.372 15114 INFO nova.compute.manager [req-560be7dc-1c0c-454e-9100-acfaf9a533b4 None None] Updating bandwidth usage cache
...
2014-05-28 11:31:39.776 15114 INFO nova.compute.manager [-] Updating bandwidth usage cache
...
2014-05-28 11:32:40.232 15114 INFO nova.compute.manager [-] Updating bandwidth usage cache


With the patch that only happens once:

2014-05-28 11:34:06.097 15304 INFO nova.compute.manager [-] Updating bandwidth usage cache
2014-05-28 11:34:06.126 15304 WARNING nova.compute.manager [-] Bandwidth usage not supported by hypervisor.
2014-05-28 11:35:06.156 15304 AUDIT nova.compute.resource_tracker [-] Auditing locally available compute resources
Comment 6 errata-xmlrpc 2014-05-29 20:35:36 UTC 
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

http://rhn.redhat.com/errata/RHSA-2014-0578.html
Note You need to log in before you can comment on or make changes to this bug.