Bug 1489120 - Ceph pg_num and pgp_num default is too low for any user
Summary: Ceph pg_num and pgp_num default is too low for any user
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Red Hat OpenStack
Classification: Red Hat
Component: openstack-tripleo-heat-templates
Version: 11.0 (Ocata)
Hardware: x86_64
OS: Linux
high
high
Target Milestone: z4
: 11.0 (Ocata)
Assignee: Giulio Fidente
QA Contact: Yogev Rabl
URL:
Whiteboard:
Depends On: 1481821 1489121
Blocks:
TreeView+ depends on / blocked
 
Reported: 2017-09-06 17:58 UTC by jomurphy
Modified: 2018-02-13 16:31 UTC (History)
9 users (show)

Fixed In Version: openstack-tripleo-heat-templates-6.2.4-2.el7ost
Doc Type: Bug Fix
Doc Text:
Prior to this update, the default `pg_num` and `pgp_num` values set for the Ceph pools (32) was too low. For most production environments, using only 32 PGs for a Ceph pool could have resulted in performance issues, as it limited the scalability of the writes on multiple OSDs, and increased the time necessary to recover from faults. With this update, the new `pg_num` and `pgp_num` default for the Ceph pools is set to 128. As a result, the new default will result in improved Ceph cluster performance when there are 5-10 OSDs, but it is always best to calculate and optimize the PGs allocated to each pool for the different use cases. See the PGs chapter in the Ceph product documentation for information on how to calculate an optimal value: https://access.redhat.com/documentation/en-us/red_hat_ceph_storage/2/html/storage_strategies_guide/placement_groups_pgs.
Clone Of: 1481821
Environment:
Last Closed: 2018-02-13 16:31:14 UTC
Target Upstream Version:
Embargoed:

Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
OpenStack gerrit 507778 0 None stable/ocata: MERGED tripleo-heat-templates: Default Ceph pg_num and pgp_num to 128 (I1920bc8f5251f362af38ad3bd6f46dda42c6ee93) 2018-01-26 12:14:26 UTC
Red Hat Product Errata RHBA-2018:0310 0 normal SHIPPED_LIVE Red Hat OpenStack Platform 11.0 director Bug Fix Advisory 2018-02-14 00:13:02 UTC

Description jomurphy 2017-09-06 17:58:10 UTC 
+++ This bug was initially created as a clone of Bug #1481821 +++

Description of problem:
OSP director defaults set Ceph's pg_num and pgp_num to 32 which is too low. This will cause performance issues for any costumer that will use it.

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


How reproducible:
100%

Steps to Reproduce:
1.
2.
3.

Actual results:
The pg_num and pgp_num for all the pools is 32

Expected results:
pg_num and pgp_num default should be 128, which is the right number for the minimum deployment of 3 OSDs in a cluster

Additional info:

--- Additional comment from Jeff Brown on 2017-08-29 14:11:34 EDT ---

We need to set this to 128 by default. It will potentially overwritten by derived parameters or customers that  read the documentation.
Comment 3 Yogev Rabl 2018-01-25 20:59:34 UTC 
failed. 

The code has not been merged yet in the latest puddle
Comment 7 Yogev Rabl 2018-01-31 18:01:30 UTC 
Verified, the pg_num is set by default to 128
Comment 10 errata-xmlrpc 2018-02-13 16:31:14 UTC 
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2018:0310
Note You need to log in before you can comment on or make changes to this bug.