On 27/07/15 22:49, Aurelien Gonnay wrote:
Hi,
I’m currently collecting a bunch of metrics from several servers using PCP.
It has been so far a pretty convincing experience from a monitoring /
collection of metrics point of view.
That's good.
However, on the UI front, I’m having a rather frustrating experience.
When using graphite (either plain graphite or grafana) and I’m facing 2
issues:
1.Metric name encoding: our servers are using ‘-‘ in their name, which
do not play well with the metric name encoding.
Hmm ... the hyphen is in the hostname (not the metric name pe se), is
that correct?
There have been no issues (for at least 15 years!) in the core PCP code
and utilities ... I have a hosts named bozo-vm and bozo-laptop in the QA
farm and PCP is thrashed on these on a daily basis.
Now it is possible that there is something broken in pmwebd-land, but
I'll leave to the experts in those areas to comment.
...
What solutions are used by seasoned PCP users to visualize
realtime/historical metrics, for
1.dashboards used on a daily basis,
We used to have pmgadgets, but that never made the transition from SGI
gfx libraries to open source (and the configuration "language" was
Neanderthal).
[unashamed call for a volunteer willing to fix this]
2.deep-dive solution to investigate / correlate events when a given
production issue arises
My experience (only over the last 40 years) convinces me that for hard
problems, the sort of tools you need are unlikely to be web-based. Much
more insightful analysis comes from application of tools like pmie,
pmchart, pmlogsummary, pmlogdiff to the PCP archives ... these help
identify anomalies.
Another missing tool from the arsenal would be pmview for 3-D
visualization of performance scenes, but that is languishing in the
source tree, sort of 90% working (from its SGI port) and lacking someone
to give it the love it needs to make it work.
[another unashamed call for a volunteer willing to fix this]
|