Description of problem:  When running larger katello-disconnected commands (such as sort or publish with 100s of RPM repos), mongod throws this error:

"database error: too much data for sort() with no index.  add an index or specify a smaller limit"

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

pulp-katello-0.5-1.el7sat.noarch
ruby193-rubygem-katello-2.2.0.76-1.el7sat.noarch
katello-utils-2.2.5-1.el7sat.noarch

How reproducible: Usually


Steps to Reproduce:
1.  Import a few hundred large RPM repositories with katello-disconnected.
2.  Run a katello-disconnected publish.
3.  Watch journalctl throw mongod and katello errors similar to above.

Actual results:

"database error: too much data for sort() with no index.  add an index or specify a smaller limit"


Expected results:

No error, complete publishing of all RPMS.

Additional info:

This fixes the issue.  Perhaps it should be added to setup phases of katello (migrate DB)?

[root@host ~]# mongo
MongoDB shell version: 2.4.9
connecting to: test
> use pulp_database
switched to db pulp_database
> db.units_rpm.ensureIndex( { name: 1, epoch: 1, version: 1, release: 1, arch: 1, checksumtype: 1, checksum: 1 } )