Bug 794885 - configure mongodb for numa machine
Summary: configure mongodb for numa machine
Keywords:
Status: CLOSED WONTFIX
Alias: None
Product: CloudForms Cloud Engine
Classification: Retired
Component: aeolus-configure
Version: 1.0.0
Hardware: Unspecified
OS: Unspecified
medium
medium
Target Milestone: rc
Assignee: Mike Orazi
QA Contact: wes hayutin
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2012-02-17 20:44 UTC by Dave Johnson
Modified: 2013-01-16 18:26 UTC (History)
5 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2013-01-16 18:26:57 UTC


Attachments (Terms of Use)

Description Dave Johnson 2012-02-17 20:44:06 UTC
Description of problem:
===============================
Not sure how big of a deal this is.  Basically I deployed a application earlier today and came back to the server hours later to deploy again and found mongodb/warehouse were stopped.

Looking through the logs, I found that mongo seemed to have crashed.  Not sure if that is related to it being a numa machine but if mongo needs a special configuration for numa, we probably need to detect and do the recommended configuration.

Copying warehouse guys for input...


Additional info:
=======================

numa = Non-Uniform Memory Access

snippet from mongodb.log
----------------------------- 
Fri Feb 17 12:18:45 [initandlisten] MongoDB starting : pid=3083 port=27017 dbpath=/var/lib/mongodb 64-bit
 
** WARNING: You are running on a NUMA machine.
**          We suggest launching mongod like this to avoid performance problems:
**              numactl --interleave=all mongod [other options]
 
Fri Feb 17 12:18:45 [initandlisten] db version v1.8.2, pdfile version 4.5
Fri Feb 17 12:18:45 [initandlisten] git version: nogitversion
Fri Feb 17 12:18:45 [initandlisten] build sys info: Linux hs20-bc2-5.build.redhat.com 2.6.18-274.17.1.el5 #1 SMP Wed Jan 4 22:45:44 EST 2012 x86_64 BOOST_LIB_VERSION=1_41
**************
old lock file: /var/lib/mongodb/mongod.lock.  probably means unclean shutdown
recommend removing file and running --repair
see: http://dochub.mongodb.org/core/repair for more information
*************
Fri Feb 17 12:18:45 [initandlisten] exception in initAndListen std::exception: old lock file, terminating
Fri Feb 17 12:18:45 dbexit:
Fri Feb 17 12:18:45 [initandlisten] shutdown: going to close listening sockets...
Fri Feb 17 12:18:45 [initandlisten] shutdown: going to flush diaglog...
Fri Feb 17 12:18:45 [initandlisten] shutdown: going to close sockets...
Fri Feb 17 12:18:45 [initandlisten] shutdown: waiting for fs preallocator...
Fri Feb 17 12:18:45 [initandlisten] shutdown: closing all files...
Fri Feb 17 12:18:45 closeAllFiles() finished
Fri Feb 17 12:18:45 dbexit: really exiting now

Comment 1 Dave Johnson 2012-02-17 20:52:12 UTC
Mongo is now in a bad state and not able to restart (aeolus-restart-services) without manually removing old lock file (/var/lib/mongodb/mongod.lock)

Comment 2 Mike Orazi 2012-02-21 15:34:04 UTC
Deploying to numa isn't on scope for 1.0, but we'll look at it in the future.

Comment 4 Mike Orazi 2013-01-16 18:26:57 UTC
New version of imagefactory does not require iwhd, which resolves the dep on mongodb.


Note You need to log in before you can comment on or make changes to this bug.