multi-line logs are difficult to handle, it would be far easier on you if you can turn them into single-line logs as early in processing as possible.

There is a lot of business analytics value in logs. the 'easy' way is to throw it into Splunk or ElasticSearch and depend on queries there, but that ends up being rather inefficient. I like to get the logs into those tools to make them easy to explore, but once you figure out what you want to know you can be far more efficient in the gathering of your metrics.

you can use something like Simple Event Correlator to turn a series of events into counts that you can then graph, and once you have graphable numbers, then something like the holt-winters algorithm that RRDtool implements can predict normal values and alert you when you stray (and the beauty of holt-winters is that the same numerical value can produce a 'unexpecteedly high' alert at 3am sunday morning, 'unexpectedly low' at 10am monday, and be in the normal range at 3pm on monday)

Rsyslog is not an analysis engine, but it's a very good routing/reformating engine for single-line logs (it can do some handling of multi-line logs, but that tends to just push the failure down to the next component)

One thing to remember is that rsyslog is a 'best effort' logging, there are ways to make it handle failures, but there remain failures that can cause logs to be lost. Don't use rsyslog as the only path for content that will cost you money if it's lost.

https://www.usenix.org/publications/login/david-lang-series
https://www.usenix.org/publications/login/april14/lang
https://www.usenix.org/conference/lisa12/technical-sessions/presentation/lang_david
http://ristov.users.sourceforge.net/publications/cogsima15-sec-web.pdf

David Lang



On Fri, 9 Jul 2021, Jim Van Meggelen via rsyslog wrote:

Date: Fri, 9 Jul 2021 07:42:28 -0500 (CDT)
From: Jim Van Meggelen via rsyslog <[email protected]>
To: rsyslog-users <[email protected]>
Cc: Jim Van Meggelen <[email protected]>
Subject: Re: [rsyslog] using Kibana / OpenSearch Dashboards to analyze logs
    during development

Daniel,

I'm pretty sure you and I have had at least one yap at some conference or 
another. Could be I just attended a talk of yours.

I saw your name here and thought "I'm pretty sure I've met him somewhere", and 
that was somewhat of a pleasant shock, because I've been digging into rsyslog for some 
stuff I've been thinking about, and it's in a similar vein to what you're talking about 
here (feeling multi-line data into analytics to help make some sense of it), and frankly 
it's nice to hear someone else in the same line of work is thinking similar things with 
respect to these log files (which are chock full of detailed data).

I don't know if what we're after is in fact the same (most folks seem to use 
logging for error handling, whereas I'm thinking more about gleaning business 
analytics from the data).

It feels like there's gold in all those log files. It'd be interesting to see 
how it could be mined.

Regards,

Jim



_______________________________________________
rsyslog mailing list
https://lists.adiscon.net/mailman/listinfo/rsyslog
http://www.rsyslog.com/professional-services/
What's up with rsyslog? Follow https://twitter.com/rgerhards
NOTE WELL: This is a PUBLIC mailing list, posts are ARCHIVED by a myriad of 
sites beyond our control. PLEASE UNSUBSCRIBE and DO NOT POST if you DON'T LIKE 
THAT.

Reply via email to