Bug 489812 - rhnpushing content into a channel gets *really* slow
Summary: rhnpushing content into a channel gets *really* slow
Keywords:
Status: CLOSED NOTABUG
Alias: None
Product: Spacewalk
Classification: Community
Component: Server
Version: 0.5
Hardware: All
OS: Linux
low
medium
Target Milestone: ---
Assignee: Brad Buckingham
QA Contact: Red Hat Satellite QA List
URL:
Whiteboard:
Depends On:
Blocks: space05
TreeView+ depends on / blocked
 
Reported: 2009-03-11 23:41 UTC by Mike McCune
Modified: 2012-03-06 08:06 UTC (History)
1 user (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2009-03-16 16:47:53 UTC
Embargoed:


Attachments (Terms of Use)

Description Mike McCune 2009-03-11 23:41:25 UTC
* Create a rhel5-i386 channel in your Spacewalk 0.5 server

* Mount engarchive:

in /etc/fstab:

/vol/engineering/devarchive/redhat /mnt/redhat    nfs   defaults   0 0

* cd /mnt/engarchive/released/RHEL-5-Server/U3/i386/os/Server

* rhnpush -vvv --server=localhost --user=<user> --password=<pass> --channel=rhel5-i386 --dir=/mnt/engarchive/released/RHEL-5-Server/U3/i386/os/Server

Now it will start off fast and start pushing files after it does the MD5 sum but then it starts dragging bad ... and Oracle takes up most of the CPU.

I know that Prad reworked the transactions around how we commit during rhnpush but this may have made rhnpush slower.  Often times it just appears to sit there and do nothing between packages for up to 60-120 seconds.

I know Prad is out but this seems like a nasty regression from 0.4 and should try and get it looked at before 0.5 goes out.

Comment 1 Brad Buckingham 2009-03-12 21:07:27 UTC
I performed an rhnpush of the rhel5-i386 packages (2255 packages) on fjs-0-03 using:

1. space04 installation using the official 0.4 distribution
2. space05 installation using the nightly build from miroslav

In each test, I performed a pxe provision of fjs-0-03 to RHEL 5.3 prior to installing Spacewalk.

Observations:

1. rhnpush on space05 actually completed in less time than on space04

    space04:

    real    108m8.982s
    user    1m25.087s
    sys     0m37.990s

    space05: 

    real    69m44.451s
    user    1m25.442s
    sys     0m36.987s

2. during each rhnpush, did observe the cpu utilization of the oracle process reach 90+% on several occasions

3. during each rhnpush, did observe that some packages did take a long time to push (e.g. ~60 sec)...   a check on a several these, observed that it occurred on packages that were much larger in size (e.g. 50MB)

Comment 2 Mike McCune 2009-03-16 16:47:53 UTC
awesome, thanks for testing!  closing as NOTABUG


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