Status and suggestion

131 views
Skip to first unread message

Scott H.

unread,
Mar 17, 2016, 1:19:35 PM3/17/16
to metrics2.0
I would like to know what the status of this effort is?

Also I would like to suggest that rather than using "server" and "device" that you move to "device_name" and "device_type", this would allow one to post metrics for switches/routers/appliances etc.  And use the same fields for servers.

Scott

Dieter Plaetinck

unread,
May 5, 2016, 7:23:19 PM5/5/16
to Scott H., metrics2.0
Hi Scott,
progress has been stagnant and people have a hard time adopting, never mind contributing. a few reasons come to mind:
1) i have failed to make the spec clear in what it covers and what it doesn't, making it hard for people to adopt or see what's there to adopt. Also we should untie the spec better from some of the implementations i did for the graphite stack, none of the graphite pieces belong in the metrics2.0 specification.
2) the model of a metric identified as nothing but a set of tags is too weird for many to swallow. it should be adjusted for tags to just augment some kind of identifier value (which could be a string, some people call it a key or 'name' property)
3) i have personally had not much time after leaving vimeo and joining raintank (http://dieter.plaetinck.be/post/focusing-on-open-source-monitoring-joining-raintank/) to work on metrics2.0 , been busy with getting the basics of our monitoring stack working, but now I'm finding some more time


re your suggestion:  device is quite generic and commonly also used for devices on a host (such as NIC, block device etc), so how do you tag the series that represents a given storage device on a particular host? with the current model this would be {host:my-server device=/dev/sda } with your suggestion it seems both device_name and device_type would need to occur twice (once to describe the server we're on, once to talk about the storage device on that server) which seems complicated to come up with a model for, as well as an implementation. or i guess we would only be allowed to use device_name and device_type for the "parent" device (server/switch/...) and any internal devices like blockdevices would get another tag like blockdevice=/sda or nic=eth0 etc which seems to defeat your goal.

btw there's a PR for server to host : https://github.com/metrics20/spec/pull/3





--
You received this message because you are subscribed to the Google Groups "metrics2.0" group.
To unsubscribe from this group and stop receiving emails from it, send an email to metrics20+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply all
Reply to author
Forward
0 new messages