Enhanced parser

[rendicott]

Love the project by the way. However, a lot of my perfdata comes from windows boxes and NSClient++ which spits back nasty perfdata UOM strings. raws[0] and raws[1] in the below example are actual sample data that makes up 80% of my disk space checks. These were all throwing exceptions in the process_log() function because it was splitting on spaces. I wrote a parser_enhanced() function that finds the true delimiter with nasty perfdata strings like this.

Let me know if it looks like it could benefit the project.

raws = [    "c: free=201.06451GB;0.97656;0.48828;0;279.37899 c: free %=71%;0;0;0;100",
            "total 5m=2%;80;90 total 1m=1%;80;90 total 5s=2%;80;90",
            "my perf/descrip is awsom!=20gbs;30ms;70rs dis$isdope=45%;80gs",
]

Types of exceptions I was seeing

How it was showing up in Graphite:

[shawn-sterling]

Hi There, sorry for the delay I was at a conference (re:invent) all week.

Thank you for taking the time to write this.

Can you take a peek at the 'parserfix' branch and see if that works with your data? The branch is going to be a bit behind now, because for some reason people keep saying they will go test it and no-one ever gets back to me. :) heh

Anyways, if the parserfix branch doesn't fix the problem, then I'll take a closer look at this and we'll move forward with it.

Have a good day.

[rendicott]

Hey Shawn,
I took a look at the 'parserfix' branch and I see how it would handle labels with a single quote in the name since it's doing a shlex.split() instead of the standard string split(). However, even with the shlex.split() labels with spaces in the name will still be a problem.

e.g.,

>>> import shlex
>>> shlex.split(s)
['c:', 'free=201.06451GB;0.97656;0.48828;0;279.37899', 'c:', 'free', '%=71%;0;0;0;100']

It looks like that's all that the parserfix branch is doing is those few lines.

I checked to make sure my parser could handle labels with single quotes in them as was intended with the shlex split using this raw data:

raws = [    "c: free=201.06451GB;0.97656;0.48828;0;279.37899 c: free %=71%;0;0;0;100",
            "total 5m=2%;80;90 total 1m=1%;80;90 total 5s=2%;80;90",
            "my perf/descrip is awsom!=20gbs;30ms;70rs dis$isdope=45%;80gs",
            "'label cool'=56mb;10mb;8mb bob's awesome label=45mbps;43mbps",
]

C:\source\rendicott-github\graphios2>python enhancedparser.py
CHUNKS= '['c:_free=201.06451GB;0.97656;0.48828;0;279.37899', 'c:_free_%=71%;0;0;0;100']'
CHUNKS= '['total_5m=2%;80;90', 'total_1m=1%;80;90', 'total_5s=2%;80;90']'
CHUNKS= '['my_perf/descrip_is_awsom!=20gbs;30ms;70rs', 'dis$isdope=45%;80gs']'
CHUNKS= '["'label_cool'=56mb;10mb;8mb", "bob's_awesome_label=45mbps;43mbps"]'

It looks like it handles it just fine. My new function in the enhanced-parser branch should be able to handle any character in the label name as long as it's not an = or ; which I think is similar to how pnp4nagios handles parsing.

Please do test it though :-)