Project

General

Profile

Actions
Copy link

Support #8826

closed

Attempt to set PG_NUM and PGP_NUM to 8192 on pool rbd causes OSDs to go dow after

Added by Jean-Charles Lopez almost 10 years ago. Updated almost 10 years ago.

Status:
Rejected
Priority:
High
Assignee:
-
Category:
-
Target version:
-
% Done:

0%

Tags:
Reviewed:
Affected Versions:
Pull request ID:

Description

The customer is willing to modify pool rbd for its production usage but doing so generates an error message when issuing the ceph osd pool set rbd pg_num 8192 and causes many OSDs to go down.

The zendek issue is https://inktank.zendesk.com/agent/#/tickets/1655

I have attached the information I collected while discussing with the customer

Customer has installed 0.80.3 on CentOS 6.5

JC

Files

Download all files
ceph_health_detail.txt (148 KB) ceph_health_detail.txt Jean-Charles Lopez, 07/14/2014 07:51 AM
ceph_osd_dump.txt (36 KB) ceph_osd_dump.txt Jean-Charles Lopez, 07/14/2014 07:51 AM
ceph-s.txt (980 Bytes) ceph-s.txt Jean-Charles Lopez, 07/14/2014 07:51 AM
ceph.conf.txt (493 Bytes) ceph.conf.txt Jean-Charles Lopez, 07/14/2014 07:51 AM
CrushMap.txt (7.42 KB) CrushMap.txt Jean-Charles Lopez, 07/14/2014 07:51 AM
HW_Config.txt (207 Bytes) HW_Config.txt Jean-Charles Lopez, 07/14/2014 07:51 AM
LastCephOsdTree.txt (7.75 KB) LastCephOsdTree.txt Jean-Charles Lopez, 07/14/2014 07:51 AM
osd-0.log.gz (280 KB) osd-0.log.gz Jean-Charles Lopez, 07/14/2014 07:51 AM
ceph.log.txt (1.5 MB) ceph.log.txt Jean-Charles Lopez, 07/14/2014 07:51 AM
Actions
Copy link
 #1

Updated by Greg Farnum almost 10 years ago

  • Description updated (diff)
Actions
Copy link
 #2

Updated by Greg Farnum almost 10 years ago

  • Status changed from New to Rejected

Based on other analysis in the private ticket, it looks like it's just hitting the fd limit; I think that's well-documented elsewhere and it's not anything we're going to solve except by rewriting the messenger.

Actions
Copy link

Also available in: Atom PDF