Yesterday, we changed the way we access our internal RT server. We used to
go with http://rt, but now we go with http://rt.ourdomain.com.
Since then, performance has increased tremendously! I mean, tickets that
used to take 10-15 seconds to open, now open in 3-4 seconds.
Has anyone experienced this?
We host our own (dual server) DNS and “RT” is an alias (CNAME) for the
server it runs on (it’s not a separate server). Apache is set up as
follows:
ServerName rt
ServerPath /rt
Options -Indexes
TransferLog /var/log/httpd/rt.access_log
ErrorLog /var/log/httpd/rt.error_log
UserDir disabled
ScriptAlias /cgi-bin/ /home/httpd/prodcgi-bin/rt/
ServerAlias rt
From a DNS perspective, and Apache’s use of it, we don’t see any reason for
the increase in performance.
Our domain DOES have a Squid proxy and that proxy runs on the same server RT
runs (“Aha!”, you say, “That’s it!”). HOWEVER, we have the browsers set to
bypass the proxy for internal servers. Plus, looking at the cache’s stats,
after hours us constant use today, the only RT object cached is:
GET http://rt.ourdomain.com/NoAuth/images//bplogo.gif
SQUID is set to NOT cache “query” URLs (e.g.,
http://xyz.abc.com/cgi-bin/cgiprog?query=here).
Also, the 3-4 second ticket open time was measured after clearing the Squid
and browser caches.
Are we missing something obvious here? If not, could using the alias name
(instead of the FQDN) really be affecting performance that much?
Basic stats:
RT 3.0.8
Perl 5.8.0
MySQL 3.23.58
Chago