pcp
[Top] [All Lists]

Re: [pcp] proc PMDA problems

To: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Subject: Re: [pcp] proc PMDA problems
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon, 4 May 2015 21:12:40 -0400 (EDT)
Cc: PCP <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <5547D742.4060807@xxxxxxxxxxxxxxxx>
References: <5547D742.4060807@xxxxxxxxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: odiaXvkC5XJshXzVRd+pEEfHckgnZw==
Thread-topic: proc PMDA problems

----- Original Message -----
> [...]
> 
> In both cases, since I did not specify any instances, I would have
> expected "No data available" or a reason why no data was available.

Yep.  So the proc.psinfo.cpusallowed has behaved this way for some
time, and I followed that pattern for the recent proc.psinfo.ngid
addition.  Both are giving an error due to lack of kernel support
on those platforms.

I agree with you though, its the wrong error code - I'll update it.
We use PM_ERR_APPVERSION elsewhere in the PMDA for this, so I'll go
with that.

> I'm also seeing qa/999 failures with lines like:
> 
> hotproc.psinfo.cgroups: No data available
> 
> I suspect this one is OK, but just needs filtering in qa/999 ... does
> that sound right?

Yep.  We actually use a mix of -ENODATA (entire procfs file missing, or
some other IO error) and PM_ERR_APPVERSION (individual fields of procfs
file missing).

cheers.

--
Nathan

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