pcp
[Top] [All Lists]

Re: [pcp] zbxpcp: add derived metrics support

To: Nathan Scott <nathans@xxxxxxxxxx>
Subject: Re: [pcp] zbxpcp: add derived metrics support
From: Marko Myllynen <myllynen@xxxxxxxxxx>
Date: Tue, 16 Feb 2016 19:44:10 +0200
Cc: pcp developers <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <561951755.20107378.1455148828758.JavaMail.zimbra@xxxxxxxxxx>
Organization: Red Hat
References: <56BAF517.4030300@xxxxxxxxxx> <561951755.20107378.1455148828758.JavaMail.zimbra@xxxxxxxxxx>
Reply-to: Marko Myllynen <myllynen@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.5.0
Hi,

On 2016-02-11 02:00, Nathan Scott wrote:
> ----- Original Message -----
>>
>> Turns out adding derived metrics support to zbxpcp is trivial.
>>
>> Based on comments from Ken yesterday I think we can ignore errors from
>> pmLoadDerivedMetrics, if the configuration file is malformed then
>> messages will be logged into the zabbix-agent log file.
>>
>> Update the man page accordingly. The file name is quite a name monster,
> 
> (seems OK to me - leaves no doubt in the mind of the reader as to what it
> is for, anyway)
> 
>> please feel free to adjust if you can come up with something better.
>>
>> A small CodingStyle fixlet included as a bonus as well.
> 
> Looks good - warrants a tweak to qa/1067 to regression test it?

Metrics with instances wouldn't work if the style fix would break things
so I think we're ok in that sense but qa/1067 may need a bit of other
kind of tweaking, I'm seeing:

1067 - output mismatch (see 1067.out.bad)
123,125c123,125
< pcp.sample.dynamic.counter[one]               VALUE
< pcp.sample.dynamic.discrete[one]              VALUE
< pcp.sample.dynamic.instant[one]               VALUE
---
> pcp.sample.dynamic.counter[1]                 VALUE
> pcp.sample.dynamic.discrete[1]                VALUE
> pcp.sample.dynamic.instant[1]                 VALUE

So the results depend on the contents of the
/var/lib/pcp/pmdas/sample/dynamic.indom file (and that it actually
exists, the failure is different it doesn't). Do you think it's
worthwhile to keep these metrics as part of 1067? If so, how to make the
test tolerant to missing / different dynamic.indom files?

Thanks,

-- 
Marko Myllynen

<Prev in Thread] Current Thread [Next in Thread>