Bug 1014663 - Content View publish takes 2-10x as long with MDP2
Summary: Content View publish takes 2-10x as long with MDP2
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: Pulp
Classification: Retired
Component: rpm-support
Version: Master
Hardware: Unspecified
OS: Unspecified
urgent
high
Target Milestone: ---
: 2.3.0
Assignee: Barnaby Court
QA Contact: pulp-qe-list
URL:
Whiteboard:
Depends On: 1014390
Blocks: 950743 1017359
TreeView+ depends on / blocked
 
Reported: 2013-10-02 13:49 UTC by Mike McCune
Modified: 2013-12-09 14:30 UTC (History)
4 users (show)

Fixed In Version:
Clone Of: 1014390
: 1017359 (view as bug list)
Environment:
Last Closed: 2013-12-09 14:30:27 UTC
Embargoed:


Attachments (Terms of Use)

Description Mike McCune 2013-10-02 13:49:56 UTC
+++ This bug was initially created as a clone of Bug #1014390 +++

Content View publishing of a moderate sized repo (3k packages, RHEL 6.4-x86_64) takes 20+ minutes to publish the first content view in MDP2

2013-10-01T16:07:48-0400: [Worker(delayed_job host:dhcp-8-30-94.lab.eng.rdu2.redhat.com pid:18493)] ContentViewDefinition#generate_repos completed after 1328.9986

With MDP1 it takes around 3 minutes:

2013-09-08T12:11:41-0400: [Worker(delayed_job host:satellite6.usersys.redhat.com pid:2493)] ContentViewDefinition#generate_repos completed after 372.8071



Steps to reproduce:

1) Enable RHEL 6.4 x86_64 from Red Hat repositories
2) Sync
3) create a content view with only this one repo in it, no filters
4) publish

wait 20 minutes ...

We are touting MDP2 to be faster with CV publishing and it doesn't appear to be so.

--- Additional comment from RHEL Product and Program Management on 2013-10-01 17:17:44 EDT ---

Since this issue was entered in Red Hat Bugzilla, the pm_ack has been
set to + automatically for the next planned release

--- Additional comment from David Davis on 2013-10-02 09:14:08 EDT ---

Publishing a definition with only two small zoo repos is taking about 100 seconds. 93 seconds are being spent in ContentViewDefinition#generate_repos. Here's a breakdown of the times.

32 seconds
https://github.com/Katello/katello/blob/cc81e688b0d01deb19b6e85a7d5d094ad17e186e/app/models/content_view_definition.rb#L292-L294

34 seconds
https://github.com/Katello/katello/blob/cc81e688b0d01deb19b6e85a7d5d094ad17e186e/app/models/content_view_definition.rb#L311-L312

23 seconds
https://github.com/Katello/katello/blob/cc81e688b0d01deb19b6e85a7d5d094ad17e186e/app/models/content_view_definition.rb#L89

95% of the time here is spent waiting on Pulp.

--- Additional comment from David Davis on 2013-10-02 09:19:11 EDT ---

Here is the pulp log:

https://gist.github.com/daviddavis/a8d6bcc8ce82afe489da

--- Additional comment from David Davis on 2013-10-02 09:22:47 EDT ---

I'm using pulp 2.3.0-0.12.alpha but I believe MDP2 may be on 2.3.0-0.14.alpha.

Comment 2 David Davis 2013-10-03 20:36:05 UTC
So I found out the wait times were not because of pulp but because of the code in the method we use to poll pulp tasks' statuses. By changing the sleep time, we were able to reduce the times for the repos in comment #1 from 100+ seconds to less than 30 seconds.

Comment 3 Barnaby Court 2013-10-04 18:25:45 UTC
Partha verified that running with a larger repo (rhel) synced in a smaller amount of time.  This report is being treated as an anomaly and if get further reports we will re-open it.

Comment 4 Preethi Thomas 2013-10-04 19:08:24 UTC
Moving to verified as per the above comments.

Comment 5 Preethi Thomas 2013-12-09 14:30:27 UTC
Pulp 2.3 released.


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