Bug 1047872

Description of problem:

The mongodb storage driver currently uses an aggregation pipeline over the meter collection in order to construct a list of resources adorned with first & last sample timestamps etc.

The problem with this approach is that the mongodb aggregation framework performs sorting in-memory, in this case operating over a potentially very large collection (particularly if the GET /v2/resources was not constrained with query params, e.g. to limit to a single tenant for example).

It turns out the mongodb innards are hardcoded to abort any sorts in an aggregation pipeline that will consume more than 10% of physical memory.


Version-Release number of selected component (if applicable):

mongodb-server-2.4.6-1.el6.x86_64
openstack-ceilometer-api-2013.2-4.el6ost.noarch
openstack-ceilometer-central-2013.2-4.el6ost.noarch
openstack-ceilometer-collector-2013.2-4.el6ost.noarch
openstack-ceilometer-common-2013.2-4.el6ost.noarch


How reproducible:

100% if the meter collection is sufficiently large.


Steps to Reproduce:

1. Allow meter collection to grow to at least X elements (actual value of X to be filled in by gilles, who has observed this issue in production with the new internal lab).

Note that the meter collection size can be retrieved via:

  $ mongo ceilometer
  > db.meter.count()


2. Attempt to list resources with an unconstrained query:

  $ ceilometer resource-list


Actual results:

The resource listing fails:

  $ ceilometer resource-list
  WARNING (http:172) Request returned failure status.
  HTTPInternalServerError (HTTP 500)

with an error similar to the following observed in the API logfile /var/log/ceilometer/api.log:

2013-12-17 03:56:57.516 21917 ERROR wsme.api [-] Server-side error:
"command SON([('aggregate', u'meter'), ('pipeline', [{'$match': {}},
{'$sort': {'timestamp': -1, 'project_id': -1, 'user_id': -1}},
{'$group': {'meters_unit': {'$push': '$counter_unit'}, 'source':
{'$first': '$source'}, 'project_id': {'$first': '$project_id'},
'user_id': {'$first': '$user_id'}, 'last_sample_timestamp': {'$max':
'$timestamp'}, 'meters_name': {'$push': '$counter_name'},
'first_sample_timestamp': {'$min': '$timestamp'}, 'meters_type':
{'$push': '$counter_type'}, '_id': '$resource_id', 'metadata':
{'$first': '$resource_metadata'}}}])]) failed: exception: terminating
request:  request heap use exceeded 10% of physical RAM".
Detail:<TRUNCATED>


Expected results:

The resource list should display all known resources.


Additional info:

The issue can worked around by partitioning the resource query per-tenant, e.g.:

  for project in $(keystone tenant-list | awk '/ True / {print $2}')
  do
    ceilometer resource-list -q project=$project | grep -vE '(\+-|Resource ID)'
  done