D 2020-12-16T16:53:59.813
L logfmt1
N text/x-markdown
P b50bde087462897cce397825b6c409d3a170baf90ae85e7a7758aa3b39432e6d
U mario
W 2849
**logfmt1** is meant for universal log parsing, whilst reducing manual
configuration or restricting to basic log variants. It handles `*.log.fmt`
files to transform LogFormat / placeholder strings to regular expressions
(with named capture groups).
{
"class": "apache combined",
"record": "%h %l %u %t \"%r\" %>s %b",
}
For instance would resolve to:
(?<remote_host>[\\w\\-.:]+) (?<remote_logname>[\\w\\-.:]+) (?<remote_user>[\\-\\w@.]+)
\\[?(?<request_time>\\d[\\d:\\w\\s:./\\-+,;]+)\\]? "(?<request_line>(?<request_method>\\w+)
(?<request_path>\\S+) (?<request_protocol>[\\w/\\d.]+))" (?<status>-|\\d\\d\\d)
(?<bytes_sent>\\d+|-)'
This python package currently just comes with:
* `.fmt` definitions for apache + strftime + grok placeholders.
* `logex` - a basic log extractor
* And `update-logfmt` to create/rewrite `*.log.fmt` files globally.
It originated in [modseccfg](https://fossil.include-once.org/modseccfg/).
You should ideally install the [system package](https://apt.include-once.org/)
however:
apt install python3-logfmt1
This will yield the proper `/usr/share/logfmt/` structure and the run-parts
wrapper `update-logfmt`.
### logfmt1
To manually craft a regex:
import logfmt1, json
fmt = json.load(open("/.../access.log.fmt", "r"))
rx = logfmt1.regex(fmt)
rx = logfmt1.rx2re(rx) # turn into Python regex
Or with plain old guesswork / presuming a standard log format:
rx = logfmt1.regex({"class": "apache combined"})
Though that's of course not the intended use case, and hinges on
predefined formats in /usr/share/logfmt/.
### logfmt1.logopen()
`logopen(fn=…)` is basically a file-like iterator that yields
dictionaries rather than text strings.
for row in logfmt1.logopen(".../access.log"):
print(row["request_time"])
And it provides a basic regex/formatstring debugging feature (via
`debug=True` parameter or with `logex -D`):
### logex
Very crudementary extractor for log files:
logex .../access.log --tab @host @date +id
Which also handles the `.fmt` implicitly. (Kinda the whole point of
this project.)
### update-logfmt
The Python package does bundle a run-parts wrapper, but just the apache
collector, and a local Python copy of the format database. It should discover
all (Apache) `*.log` files nonetheless and pair them with `.fmt` declarations.
And that's sort of the main aspect of this project. Establish .log.fmt files
until application vendors come around to making logs parseable. The rules
database structure is subject to change, and only one possible implementation.
There might also be simpler approaches (grok mapping) to generate regexps
for format strings.
Z 88faf714b73ee7b9cb88d91c87bee093