yahoo-eng-team team mailing list archive

[Max <2063342@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

Description
===========
In a large environment the instance_actions and instance_actions_events tables for not deleted instances can grow infinitely.
For example when doing many migrations/live-migrations the events will sum up.
Every migration produces one action with ~10 events.

Huge tables in combination with the api autojoins can produce slow
queries.

Keeping migration events for one month is usually enough.

The patch in the attachment proposes a clean_events option for the nova-
manage command.


Steps to reproduce
==================
Have many VMs and instance actions e.g. live-migrations 


Expected result
===============
fast api & db queries

Actual result
=============
slow api & overloaded db due to many entries in instance_actions and
instance_actions_events - normmallymax_statement_time 

Environment
===========
nova zed
Regulary updating all hypervisors -> which means migrate VMs away 
-> instance-actions/-events tables grow

** Affects: nova
     Importance: Undecided
         Status: New

** Patch added: "0001-feat-nova-manage-db-instace-events-cleanup.patch"
   https://bugs.launchpad.net/bugs/2063342/+attachment/5770061/+files/0001-feat-nova-manage-db-instace-events-cleanup.patch

-- 
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/2063342

Title:
  slow db due to huge instance_actions tables

Status in OpenStack Compute (nova):
  New

Bug description:
  Description
  ===========
  In a large environment the instance_actions and instance_actions_events tables for not deleted instances can grow infinitely.
  For example when doing many migrations/live-migrations the events will sum up.
  Every migration produces one action with ~10 events.

  Huge tables in combination with the api autojoins can produce slow
  queries.

  Keeping migration events for one month is usually enough.

  The patch in the attachment proposes a clean_events option for the
  nova-manage command.

  
  Steps to reproduce
  ==================
  Have many VMs and instance actions e.g. live-migrations 

  
  Expected result
  ===============
  fast api & db queries

  Actual result
  =============
  slow api & overloaded db due to many entries in instance_actions and
  instance_actions_events - normmallymax_statement_time 

  Environment
  ===========
  nova zed
  Regulary updating all hypervisors -> which means migrate VMs away 
  -> instance-actions/-events tables grow

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/2063342/+subscriptions