Hi,
On 2016-06-22 04:14, Nathan Scott wrote:
> ----- Original Message -----
>>
>> Current issues:
>>
>> - pmfg rejects metrics like pmcd.pmie.configfile which pmval/pmrep/etc
>> accept
>
> Hmm, I wonder what it is about that metric it doesn't like - does it also
> reject sample.noinst? If so, that's possibly a pmfg bug - metrics can at
> any time have no values, then some values, then none again.
sample.noinst is accepted but several other metrics under pmcd are
rejected as well some under proc - looks like pmfg_lookup_item() is a
bit too strict here.
>> - in QA 1062 (vmstat-like output) idleness was previously reported
>> correctly (~99), now it get some insanely high value (115646464) which
>> gets truncated. Might also be incorrect scale parameter.
>> - QA 1068 output is different, haven't investigated yet
>> - in QA 1069 for sample.seconds we get 1 (integer) unlike with
>> pmval/pmrep which give typically something like ~0.999-1.001 and for
>> sample.milliseconds we get ~1000.002 or such where pmval/pmrep report
>> ~1.000.
>
> This may be conversion to a "time utilization" I guess - so the value
> reported will be in units of seconds-per-second - i.e. between 0 and 1
> for those metrics.
Yes, I still haven't investigated more whether I'm using incorrect scale
value or whether this is something pmfg internal.
>> - in QA 1070 earlier 0.500/0.000 is now 0 (just noticed this, haven't
>> investigated much)
>> - in QA 1071 a similar as with 1070
>
> Its great to see the QA tests being used to such good effect!
Yes, I think if we manage to sort out these issue then we can safely
assume the pmrep conversion is done correctly and then also pmfg has
been tested in real life.
Thanks,
--
Marko Myllynen
|