On 03/05/13 16:37, Marco Neumann wrote:
another logging / monitoring question is it possible to access the
source IP in the log4j conversion pattern on INFO in fuseki 0.2.6
(Build date: 2013-02-20T12:04:26+0000)?
---
Marco Neumann
KONA
It's a good one to collect.
More generally, having a standard format request log.
The IP address it can access is the immediate previous machine. But if
you're running behind httpd, it'll be 127.0.0.0 or some such, not the
originating source IP address.
IMO: If your exposing access to the public WWW, it is better to front
with locked-down reverse proxy, including connection limiting. Java web
application servers (tomcat, jetty) do need a bit of protection.
A side-effect is that you can use whatever the proxy provides for
logging, which is usually highly configurable, and it'll work with many
log analysis tools. Good place for https termination as well.
This is one reason why the Fuseki logs are highly specialised to SPARQL
- if you want request logs, there are ways to get them. Fuseki logs add
detail you can't get elsewhere or not as clearly.
What could be done is a second (optional) log, output being the usual
combined log format, and presumably incorporating X-Forwarded-For (e.g.
AWS load balancers).
http://wiki.eclipse.org/Jetty/Tutorial/RequestLog
and tomcat for a WAR version.
As a reverse proxy setup already collects this, it's not a priority for
me personally but, as ever, if anyone wants to provide a patch, it'll
get done.
Thoughts?
What I think we should avoid is bundling too much machinery into Fuseki.
Its replicating functionality available elsewhere and so it's likely
Fuseki is only going to have a not-as-good provision of such functionality.
Andy
