Feature Request: change metric name scheme . INSTANCE_BEFORE_METRIC and SEPARATE_ROLLUP
toni-moreno opened this issue · comments
Hi all.
I've been working for some years with graphite and some different metric collector. I'm surprised with the metric scheme used in vmstats.
I'm used to see metric name always under instance name, something like that.
<metric_prefix>.<hostname>.<metric_type>.<instance>.<metric_name>
by example:
production.bbdds.hostdb01.cpu.0.idle
production.bbdds.hostdb01.cpu.0.wait
In vmstats it seems like this scheme is used.
<metric_prefix>.<hostname>.<metric_type>.<metric_name>.<instance>
by example
vmstats.vcs_tag.cluster1.ESX.hostdb01.cpu.idle.27.summation
with this scheme it is difficult to see all information for each instance.
To improve this issue could be interesting to add a new configuration parameter like
INSTANCE_BEFORE_METRIC = False ( by default -- for backward compatibility)
if
INSTANCE_BEFORE_METRIC = True
The metric name :
- vmstats.vcs_tag.cluster1.ESX.hostdb01.cpu.idle.27.summation
will transform to:
- vmstats.vcs_tag.cluster1.ESX.hostdb01.cpu.27.idle.summation
Another improvement could be done if "rollup" is concatenated to the metric name , we can add a new configuration parameter like
SEPARATE_ROLLUP = True ( by default .-- for backwards compatibility--)
if
SEPARATE_ROLLUP = False
the metric name:
- vmstats.vcs_tag.cluster1.ESX.hostdb01.cpu.27.idle.summation
- vmstats.vcs_tag.cluster1.ESX.hostdb01.mem.active.average
will transform to
- vmstats.vcs_tag.cluster1.ESX.hostdb01.cpu.27.idle-summation
- vmstats.vcs_tag.cluster1.ESX.hostdb01.mem.active-average
We will avoid this gigantic tree ( see below)
Dear VmStats Users,
We stopped working on VmStats but continued to use StatsFeeder GraphiteReceiver. Please take a look at the latest GraphiteReceiver. We welcome your comments.
Thanks
Lava
@lbasavap , I will look for the StatsFeeder and GraphiteReceiver.
But I don't undestand what you mean when you say "we stopped" . do you belong to the developer group of the vmstats ?