[sldev] Jira down?

[Laurent Laborde]

On 6/25/07, Rob Lanphier <robla at lindenlab.com> wrote:
> A software upgrade is in the works.  I'm not exactly sure when the
> timing will be, but should be within the next few days.
>
> Hardware scaling is rather rough for JIRA.  It's already running on some
> fairly beefy hardware, from what I understand.  Another thing I
> understand is that clustering is not supported in JIRA, so we can't
> necessarily scale outward.  See the JIRA feature request at
> jira.atlassian.com for more on this:
> http://jira.atlassian.com/browse/JRA-7330

Holy cow ! A java application (you probably run tomcat ?) that can't
be clustered ?
i was used to manage a jserv cluster for some major phone operator, we
always found a workaround to support more load. When i left my job we
were supporting up to 2000 requests/second.
I don't like java, but if there is ONE benefit about java, it's clustering.

So ... Where is the bottleneck exactly ?
Can't you use a reverse-proxy ? Tomcat loadbalancing capabilities
(even a simple round-robin) ?
Unless the bottleneck is the database ?
How jira is mostly used ? Searching ? Read-only browsing ?

Even the frontpage take age to load... while i'm still not
authentified, and reading a page that never change. there is something
wrong here.

Wikipedia is doing very well, running 2 differents services (with a
lot of caching) for identified and unidentified users (unidentified
users are *ALWAYS* read-only, caching is good).

Also, i checked your robots.txt, you don't disallow search engine bot.
Can you check your stats and take care of thoses bots ? I know they
can generate a great amount of load on dynamic page with link,
retro-link, retro-back-link, diff, history,
post-reverse-back-retro-diff-history-link and that kind of silly
unused high load dynamic page.

-- 
kerunix Flan