yahoo-eng-team team mailing list archive

[Joe Gordon <1396677@xxxxxxxxxxxxxxxxxx>]

Even if this isn't a memory leak and not something easily fixable, the
behavior should at least be documented somewhere.

This may be relevent: http://effbot.org/pyfaq/why-doesnt-python-release-
the-memory-when-i-delete-a-large-object.htm

** Also affects: ceilometer
   Importance: Undecided
       Status: New

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1396677

Title:
  Heavy use of metering labels/rules cause memory leak in neutron server

Status in OpenStack Telemetry (Ceilometer):
  New
Status in OpenStack Neutron (virtual network service):
  New

Bug description:
  We found that large amount of metering labels and rules cause memory
  leak in neutron server. This problem is multiplied by amount of
  workers (10 workers - 10x memory leak).

  In our case we have 657 metering-lables and 122399 metering-label-
  rules,

  If anyone query them, neutron-server (worker) picked request eats
  +400Mb of memory and keep it until restart. If more requests send,
  they come to different workers cause each of them to bloat up.

  Same problem happens if neutron-plugin-metering-agent running (it send
  requests to neutron-server with same effect).

  If neutron-server hit 100% CPU  it starts to consume even more memory
  (in our case up to 1.4Gb per neutron-server worker).

To manage notifications about this bug go to:
https://bugs.launchpad.net/ceilometer/+bug/1396677/+subscriptions