Rails - Jul 2005 - lighttpd, fcgi, strangley slow

I have a rails app up and running on TxD using lighttpd/fcgi and
apache proxypass.  I am running in production mode.

However, the page takes forever to load (several seconds).  The really
strange part is that, according to my log, the rendering and db calls
are taking the expected short amount of time (faster than my local
apache/fcgi in dev mode).  But the page still takes a good several
seconds to finish loading.  In fact, it was faster in standard
apache/cgi mode than it is now in lighttpd/fcgi.

80% of the content is even coming from the (fragment) cache.  

So, according to the log, everything is fast (40+ req/sec), but the
page takes frickin forever to load.  I''m even in the same general
local as the server (TxD servers are mostly Texas, right?).

Anyone have any idea what''s going on here?

On 7/15/05, Belorion <belorion-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
wrote:
 ... snip ...
> Anyone have any idea what''s going on here?
I''ve noticed similar issues when using TERM/HUP/USR1 directly on my
dispatch.fcgi. It''s really odd and no consistent, otherwise I would
have probably bothered to report it.

Generally I''ve found that the fastest way to correct this odd behavior
is to take down lighttpd and restart it and everything gets speedy
again...

That''s my experience, at least... your mileage may vary.

Cheers,

bs.