[ZWeb] Monitoring zope.org & reducing downtime

ethan mindlace fremen mindlace@digicool.com
Tue, 05 Sep 2000 17:49:43 -0400


Jimmie Houchin wrote:
> On that site it can contain a link to a "Zope Weather" page for
> information on Zope.org's status for those who are interested.

http://www.zope.org/server-status

This gives time to fulfill request & a bunch of other useful stuff.  If
zope is handling loads of requests, it's probably just slow rather than
down.

Unfortunately it also gives out the IP addresses of people connecting,
which may be interpreted as a privacy issue.  Does anyone have thoughts
on that?

> > >1) Zope.org sits behind Apache using mod_proxy for integration.  We
> > >should find out what is the timeout for proxy connections.  If it is
> > >configurable, we should dial it _way_ down (e.g. 20 seconds).  If Zope
> > >doesn't respond, get it to say so in a reasonable period of time.

I'm not so certain on this point.  I am currently working on letting
Apache and downstream proxies cache the images & files (.tgz stuff). 
This should reduce Zope's hit rate by a _lot_.  Images and files are the
two largest things that Zope serves: with apache caching them, it should
speed everything up tremendously.

> > >2) Next, we should hack the Apache error page to be meaningful.  For
> > >instance:

If anyone knows how I trap for the "proxy not responding" error message,
I'd be happy to set something up, but I currently haven't the foggiest
idea.

> > On the "What to do when Zope.org is down" front, it really is
> > disruptive for people when they can't get the stuff they need. I know
> > long term its being fixed, but that doesn't help us between now and
> > when its nice and robust.

I should have a pager soon (in the next 2 days, i'm told), and I'm going
to let the community know its number.  I'm not sure how wide & far I'll
spread it, but at least key people will know it, and if they see a
zope.org is down message &/or notice it themselves, then they can page
me & I'll heed the call.

So your work shouldn't be slowed down for more than an hour (depending
on where I am).

The "long-term" you quote above is shorter than you might think:  By the
end of next week, there should be at least one geographically distinct
ZEO-mirror of Zope.org running.

By the end of the month (give or take) zope.org will be running on two
boxes simultaneously.  At that time, I expect very little diagnostic or
repair work that we do will take the site down, and I also expect its
general uptime to increase drastically.

does that help address things?
-- 
ethan mindlace fremen
Zopatista Community Liason
Abnegate I!