[Mongrel] Problems monitoring Mongrel with F5 BigIP

Christopher Bailey chris at codeintensity.com
Sat Aug 11 00:37:55 EDT 2007


Update...  So, I've now run Mongrel in debug mode.  I don't get any hits
from the F5.  I have ensured the Mongrel/debug mode is working by just doing
a telnet and then GET request from the F5 machine to the Mongrel, and those
requests show up in the Mongrel debug logs (and the regular Mongrel log).
So, somehow the request just isn't even getting into Mongrel.

What's odd about this, is that these exact same requests get to say the
boxes that have Nginx on them.

I also tried changing Mongrel to run on port 80 in case that was some oddity
of it, but no difference.  As Brandorr asked, the tcp and tcp half open F5
monitors are successful, just HTTP is not.

So, I will bring it all back up with F5 and have them setup a Mongrel and
test.  It appears to be something with the way an F5 makes the request or
something.  I'll also ask Jason H over at Joyent if they do direct
monitoring of Mongrels with their F5's, thanks for reminding me about that.

On 8/10/07, Brandorr <brandorr at opensolaris.org> wrote:
>
> Where did you run TCPDUMP? Or the BigIP or the Linux box? Basically I
> am wondering if the HTTP request is leaving the BigIP.
>
> Sorry to bother, I haven't had a chance to work with BigIP in aver 1
> 1/2 years, and I miss it...
>
> One other thing, Joyent and Twitter use BigIPs with Mongrel..
>
> On 8/10/07, Christopher Bailey <chris at codeintensity.com> wrote:
> > Yes, both of the standard tcp tests, and the icmp ping tests work.  We
> have
> > a slew of these machines, all configured identically, and the only case
> that
> > doesn't work is when hitting Mongrel HTTP.  Nginx works fine, as do a
> few
> > other things we have.  I will run Mongrel debug after I put the kids to
> bed
> > and see :)
> >
> >
> > On 8/10/07, Brandorr <brandorr at opensolaris.org> wrote:
> > > Christopher Bailey <chris at codeintensity.com>
> > >
> > > Do TCP open and tcp halfopen checks work
> > >
> > > On 8/10/07, Christopher Bailey < chris at codeintensity.com> wrote:
> > > > That didn't work.  I don't think you need that though.  If I just
> telnet
> > in
> > > > to the Mongrel, and do a pure "GET /heartbeat.html HTTP/1.1\n\n" it
> > works.
> > > > I am going to try running the Mongrel in debug and see, will do that
> > later
> > > > tonight...
> > > >
> > > >
> > > > On 8/10/07, Brandorr <brandorr at opensolaris.org > wrote:
> > > > > I think you need to close the connection after each health check.
> > > > > Otherwise mongrel/rails is going to block. (If my understanding is
> > > > > correct.
> > > > >
> > > > > Try the following health check and let me know how it works out:
> > > > >
> > > > > "GET / HTTP/1.1\nHost: \nConnection: Close\n
> > > > >
> > > > > On 8/10/07, Christopher Bailey <chris at codeintensity.com> wrote:
> > > > > > If this has already been covered, please point me to that (I
> didn't
> > find
> > > > > > anything in my searches)...
> > > > > >
> > > > > > We are using F5 BigIP LTM load balancers.  They have many pools
> of
> > > > Mongrels
> > > > > > they load balance across, and I of course want the F5 to know
> when a
> > > > Mongrel
> > > > > > goes down or is unavailable, etc.  To do that, I need to have an
> F5
> > > > health
> > > > > > monitor for HTTP make a request to the Mongrel.  We do this same
> > thing
> > > > with
> > > > > > our Nginx servers, and that works fine, but with Mongrel it
> doesn't.
> >  I
> > > > am
> > > > > > wondering if anyone is successfully monitoring Mongrel instances
> > from an
> > > > F5?
> > > > > >
> > > > > >
> > > > > > As further info...  I've been working with F5 tech support on
> this,
> > and
> > > > > > we're both stumped at the moment.  The same HTTP monitor works
> fine
> > with
> > > > > > Nginx.  Both try to fetch the same HTML file.  Also, if I am
> ssh'ed
> > in
> > > > to
> > > > > > the F5, and do a telnet to one of the Mongrels and do "GET
> > > > /heartbeat.html
> > > > > > HTTP/1.1", followed by two returns, this works fine.  I use that
> > same
> > > > string
> > > > > > (and two \n's) in my F5's send string for it's HTTP monitor.
> > According
> > > > to
> > > > > > F5 who looked at our tcpdump, the Mongrel just never responds
> (but
> > in
> > > > the
> > > > > > same dump, they see the direct telnet version respond
> fine).  The F5
> > > > makes a
> > > > > > direct TCP request to do this.
> > > > > >
> > > > > > Any ideas?  Suggestions?  Known issue?  Workarounds?  Anyone
> > successful
> > > > with
> > > > > > this type of setup?
> > > > > >
> > > > > > As an FYI, our system info:
> > > > > >
> > > > > > CentOS 5, nearly all up to date, 64bit, Core 2 Duo processor
> > > > > > Mongrel 1.0.1
> > > > > > Mongrel Cluster 1.0.2
> > > > > > Ruby 1.8.5 (2006-08-25) [x86_64-linux]
> > > > > >
> > > > > > F5 BigIP LTM, currently running 9.1.2 (working on moving to 9.3)
> > > > > >
> > > > > >
> > > > > > _______________________________________________
> > > > > > Mongrel-users mailing list
> > > > > > Mongrel-users at rubyforge.org
> > > > > > http://rubyforge.org/mailman/listinfo/mongrel-users
> > > > > >
> > > > > _______________________________________________
> > > > > Mongrel-users mailing list
> > > > > Mongrel-users at rubyforge.org
> > > > > http://rubyforge.org/mailman/listinfo/mongrel-users
> > > > >
> > > >
> > > >
> > > > _______________________________________________
> > > > Mongrel-users mailing list
> > > > Mongrel-users at rubyforge.org
> > > > http://rubyforge.org/mailman/listinfo/mongrel-users
> > > >
> > >
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20070810/38f66f55/attachment.html 


More information about the Mongrel-users mailing list