yahoo-eng-team team mailing list archive

[Dan Smith <dms@xxxxxxxxxxxxx>]

Public bug reported:

If a compute node references a deleted service, we will include it in
the compute node statistics output. This happens even if the compute
node record _is_ deleted, because of our join of the services table,
which causes us to get back rows anyway. This results in the stats
showing more resource than actually exists, and disagreeing with the sum
of all the individual hypervisor-show operations.

This is the query we're doing:

MariaDB [nova]> SELECT SUM(memory_mb) FROM compute_nodes JOIN services ON compute_nodes.host=services.host WHERE services.binary="nova-compute" AND compute_nodes.deleted=0;
+----------------+
| SUM(memory_mb) |
+----------------+
|        1047917 |
+----------------+
1 row in set (0.00 sec)
 
And this is what we *should* be doing

MariaDB [nova]> SELECT SUM(memory_mb) FROM compute_nodes JOIN services ON compute_nodes.host=services.host WHERE services.binary="nova-compute" AND compute_nodes.deleted=0 AND services.deleted=0;
+----------------+
| SUM(memory_mb) |
+----------------+
|         655097 |
+----------------+
1 row in set (0.00 sec)

The second value is correct for the database in question.

** Affects: nova
     Importance: Undecided
         Status: Won't Fix

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1693911

Title:
  compute node statistics will lie if service records are deleted

Status in OpenStack Compute (nova):
  Won't Fix

Bug description:
  If a compute node references a deleted service, we will include it in
  the compute node statistics output. This happens even if the compute
  node record _is_ deleted, because of our join of the services table,
  which causes us to get back rows anyway. This results in the stats
  showing more resource than actually exists, and disagreeing with the
  sum of all the individual hypervisor-show operations.

  This is the query we're doing:

  MariaDB [nova]> SELECT SUM(memory_mb) FROM compute_nodes JOIN services ON compute_nodes.host=services.host WHERE services.binary="nova-compute" AND compute_nodes.deleted=0;
  +----------------+
  | SUM(memory_mb) |
  +----------------+
  |        1047917 |
  +----------------+
  1 row in set (0.00 sec)
   
  And this is what we *should* be doing

  MariaDB [nova]> SELECT SUM(memory_mb) FROM compute_nodes JOIN services ON compute_nodes.host=services.host WHERE services.binary="nova-compute" AND compute_nodes.deleted=0 AND services.deleted=0;
  +----------------+
  | SUM(memory_mb) |
  +----------------+
  |         655097 |
  +----------------+
  1 row in set (0.00 sec)

  The second value is correct for the database in question.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1693911/+subscriptions