On 28/07/15 07:30, Frank Ch. Eigler wrote:
...
That's such a big area. I think the clinching technical complication
there is PCP's limitations in logging being governed by a static
configuration: a set of metrics and a fixed polling interval. When a
production issue arises, someone would have to notice, and reconfigure
a pmlogger instance to do more logging, and/or eyeball extra live data
interactively. It would be better if pmlogger could react dynamically.
Sorry Frank, but this is just not correct.
The mechanism exists.
pmie guard -> launch pmlc -> reconfigure pmlogger on the fly
this has been there since day 1 (almost), and has been used in anger in
real production environments with good results.
What is missing, and I believe is not achievable in a general sense, is
making this automated. It requires performance analysis and local
customization to develop the pmie guards (to start and stop detailed
logging) and to define the appropriate pmlogger config changes.
All of the PCP archive infrastructure works with archives that contain
different selections of metrics, instances and sampling intervals over
the life of the archive (subject to the usual caveats about PMDA's
playing by the rules in terms of instance domain management).
|