pcp
[Top] [All Lists]

Re: [pcp] Dynamic metric rework

To: Martins Innus <minnus@xxxxxxxxxxx>
Subject: Re: [pcp] Dynamic metric rework
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Tue, 16 Dec 2014 16:12:41 -0500 (EST)
Cc: pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <54908AB9.6080208@xxxxxxxxxxx>
References: <5481E4D7.8050700@xxxxxxxxxxx> <5489F5D1.6050208@xxxxxxxxxxx> <668428007.16198533.1418365513812.JavaMail.zimbra@xxxxxxxxxx> <548B46B3.10407@xxxxxxxxxxx> <372975962.17650087.1418622822222.JavaMail.zimbra@xxxxxxxxxx> <548F451F.7000606@xxxxxxxxxxx> <1813047828.18618124.1418729598379.JavaMail.zimbra@xxxxxxxxxx> <54908AB9.6080208@xxxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: a91pkc/apT5Ik8DQ/UcUOaXMpOcXHw==
Thread-topic: Dynamic metric rework
Hi Martins,

----- Original Message -----
> Nathan,
> 
>      OK, this QA test needs even more work.  In addition to what you
> mention below, kernel.percpu.interrupts.line1 doesn't exist on another
> machine I tested on.

Yep, I realised that later too.  :)  The kernel handles interrupts
differently between architectures & the contents of the procfs files
differs as a result.

>      The point of the test is to do a text then a fetch without any
> other calls.  So probably need to dump the tree first to get a valid
> metric to fetch, store one of those and use that as the input to the
> fetch test.

I think you can leave the test pretty much as is, if you use the
create-a-local-procfs-root-via-environment variable trick - just
create a tarball (see below qa/linux) and unpack it at the start
of the test & use the env var to point pmda_linux.so away from
the /proc of the running kernel to your new /tmp/qa-$seq-procfs.

>      I am going to try to finish up hotprocs and then come back to this,
> unless you look at it in the meantime.

It's possible, but unlikely I'll have time before I head away for some
end of year vacation time.

cheers.

--
Nathan

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