[Building Sakai] Why hasn't Recent Announcements slowed your performance?

[Matthew Jones]

I remember Michigan did have some problems with MOTD either back in 2.6 or
early 2.7 if there were a lot of MOTD's. I'm not sure if that was ever
fixed, the problem went away or if they just kept cleaning up the MOTD.
This is the same tool as announcements.

I don't ever remember any significant problems that directly pointed back
to any announcement queries, but maybe Adam does have some new details
related to the filing of that other ticket. Those queries shouldn't be
triggered by the synoptic announcements though, but would be triggered by
announcements in courses. (Where there is no date restriction on the query)

https://jira.sakaiproject.org/browse/SAK-21511<https://jira.sakaiproject.org/browse/SAK-21511?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#issue-tabs>

I'll let you know if I see anything else though. Do you have any
information about slow queries or frequent queries? Easy to generate from a
AWR report in Oracle. I'm not so sure from Mysql (probably parsing general
logs on ANNOUNCEMENT_MESSAGE).

On Tue, May 1, 2012 at 5:24 PM, Keli Sato Amann <kamann at stanford.edu> wrote:

> Thanks Matthew. Jing is going to be looking at Recent Announcements to see
> if any of the JIRAS below will help us with 2.6, as we will probably not
> upgrade versions until next summer. Julian, not sure if you'll be working
> with her, re the message date index Matthew mentions below.
>
> It does not happen all the time, but when it happens it becomes slow for
> the entire campus. I hope I'm not misquoting him, but Julian suspects that
> certain users with hundreds of active sites, or perhaps one user, might set
> this off.
>
> Keli Amann
> User Experience Specialist
> Academic Computing Services, Stanford University
>
> ----- Original Message -----
> From: "Matthew Jones" <matthew at longsight.com>
> To: "Keli Sato Amann" <kamann at stanford.edu>
> Cc: "Sakai Dev" <sakai-dev at collab.sakaiproject.org>
> Sent: Friday, April 27, 2012 2:45:35 PM
> Subject: Re: [Building Sakai] Why hasn't Recent Announcements slowed your
> performance?
>
>
> The default of the recent announcements is only to show 10 days.
>
> https://source.sakaiproject.org/svn/announcement/trunk/announcement-tool/tool/src/webapp/tools/sakai.synoptic.announcement.xml
>
>
> I can see some fixes where it might not have been respecting this limit in
> the UI pre 2.7, but you seem to say this is happening all of the time?
>
>
> https://jira.sakaiproject.org/browse/SAK-13116
> and possibly
> https://jira.sakaiproject.org/browse/SAK-12876
>
> I've never seen a significant slowdown in a 2.7+ instance related to the
> synoptic announcement box. In most instances it defaults to 10 days. At
> Michigan these were patched into our 2.6 instance almost immediately back
> in 2009.
>
>
> I'm almost entirely sure now that this field is used for the sorting,
> especially after
> https://jira.sakaiproject.org/browse/SAK-11544 (though I don't think that
> touched announcement)
>
>
> I'd maybe have your dba check to make sure the message date index is
> correct and valid. The query to return messages based on the last 10 days
> should be quick.
>
>
>
> I tested it on a DB with 20K announcements and it was instant.
>
>
> I think it was noted that the actual announcements should be cached, but I
> don't think this is related?
> https://jira.sakaiproject.org/browse/SAK-21511
>
>
> -Matthew
>
>
> On Fri, Apr 27, 2012 at 4:47 PM, Keli Sato Amann < kamann at stanford.edu >
> wrote:
>
>
> Hello
> One of the issues slowing performance on our campus is the fact that
> Recent Announcements widget queries old sites, which are not likely to have
> recent announcements. After a particularly slow day, we recently removed
> the Recent Announcements widget altogether and things seemed to be much
> better. We made a local fix to our 2.6 instance which prevented hidden
> sites from being included in search of Announcements, but few people opt to
> hide sites so we didn't have enough confidence that this alone would
> improve performance.
>
> There are a few ways we could address this so we can add that widget back
> 1) encourage users to hide sites with improved design
> 2) force course sites to be hidden after 2 years (we could do it more
> frequently, but some course admins like to refer to older sites), but make
> it more obvious how to find these hidden sites.
> 3) make course sites that are older than a quarter not show up in Recent
> Announcements search, regardless of whether they are hidden or active.
>
> Regardless of which direction we decide to go, we are surprised that this
> hasn't been a problem for other schools who have used the CLE longer and
> with more classes. We suspect it may be because other schools remove
> rosters from course sites and include new rosters every term, such that
> students don't have active sites from more than a year ago. Here, we create
> new sites for each new term a course is taught so students and instructors
> have 4 or more courses (some local admins have hundreds of courses).
> Perhaps you also encourage instructors to unpublish sites after the
> term--we don't here, because we think students like to refer to old course
> materials as they prep for new course, or as they prepare to leave and need
> to put a portfolio together, but I suppose we could force them to gather
> material before the term.
>
> If anyone who uses Recent Announcements in My Workspace has more thoughts
> on this, we'd be curious.
>
> Keli Amann
> User Experience Specialist
> Academic Computing Services, Stanford University
> _______________________________________________
> sakai-dev mailing list
> sakai-dev at collab.sakaiproject.org
> http://collab.sakaiproject.org/mailman/listinfo/sakai-dev
>
> TO UNSUBSCRIBE: send email to
> sakai-dev-unsubscribe at collab.sakaiproject.org with a subject of
> "unsubscribe"
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://collab.sakaiproject.org/pipermail/sakai-dev/attachments/20120501/61478d35/attachment.html