loki-logback-appender versus logback-access?

[rfuxx]

Hi,

in struggling with getting loki-logback-appender to work with logback-access, I have been wondering whether I am the only person trying this / being interested in this: "Directly" pushing access logs of an embedded Tomcat server into the Loki store.

Without having made a lot of attempts and analyses yet, I suppose the main problem is that...
ch.qos.logback.classic.PatternLayout extends ch.qos.logback.core.pattern.PatternLayoutBase<ILoggingEvent>
ch.qos.logback.access.common.PatternLayout extends ch.qos.logback.core.pattern.PatternLayoutBase<IAccessEvent>
But the classes of loki-logback-appender appear to expect ch.qos.logback.classic.PatternLayout, so the class hierarchy does not seem to fit together (=>ClassCastException).

We do have a working solution for all kinds of logging (and including access logging via logback-access) into the Elk Stack by making use of Logstash, and we have been requested to push the same into Loki, too (and when/if/once it works well enough, "they" will consider to get rid of the Elk Stack). So as long as I am / we are the only one(s) tying this, we might also try out the Grafana Loki Logstash Plugin with our team and whether we can reuse most of our existing logstash-logback-encoder configuration with it.

But if anyone else has been trying logback-accesss logging with loki-logback-appender, too, and maybe you have been more successful with it or have any idea how the (assumed) problems might be overcome, I would be glad to take your advice.

As might be known, logback-accesss has just recently made a major version jump after its separation from the main logback project, so it would be my/our goal to use the newer version 2.x of logback-accesss in the long term, but ideas to gain certain short-term progress with its older version 1.4.x would be appreciated, too.

Or would it take a feature request, needing explicit support for logback-access in loki-logback-appender? (Which probably takes back to the question whether I would be the only one or whether there would be others interested in it, too?)

Many thanks in advance,
Roland

[rfuxx]

Hi again,

I have meanwhile approached my challenge by trying to create a “bridge” between those parts which have in my initial attempts raised ClassCastExceptions.

When I attach the result of my attempts based on all latest-released versions here,
logstash-access-bridge-for-loki-appender.zip
you can see by the classes in com.company.logging.bridge.access which are the details of where the classes offered by the one library do not really fit with the other library, and hence the glue code is trying to “bridge” between them.

The also-included file logback-access.xml shows a way by which those “bridge” classes can be configured in a way that apparently really some access logging output reaches the Loki server.

It seems to work for the moment, but I think this solution is not very good, as it seems quite fragile against future library updates. Maybe I should not call it “bridge” but rather “quirk”.

Nevertheless it would be great if this might help in maybe finding a better approach by which the goal (to get access logs into Loki) can be reached…?

Regards
Roland