project-config/grafana/nodepool-inap.yaml
Ian Wienand d34d7ad763 Consolidate nodepool graphs
All of these dashboards are the same, and have mostly copied all the
same issues with them.  This makes updating anything a massive pain.

This implements a single dashboard template with a small script to
create individual dashboards for each provider and its regions.

I have included a range of fixes.  The y-axis format has changed in
later versions of grafana.  The API time tracking is no longer scaled,
but we just tell grafana it is in ms and it displays it correctly.
The test nodes history graph is moved to the top, as it is probably
the most interesting graph (note this splits itself out per region, if
mulitple regions are selected).  Values for "null as zero" are
consistently set.  Various formatting fixes for the labels are
included.

Change-Id: I5fbffaec3c82aa1fce0947f771de67edd15f7dfc
2018-06-22 12:21:10 +10:00

198 lines
6.1 KiB
YAML

#
# NOTE: EDIT THE TEMPLATE FILE AND RUN create-nodepool.sh
#
dashboard:
title: 'Nodepool: Inap'
templating:
- name: region
includeAll: true
multi: true
query: stats.gauges.nodepool.provider.inap-*
refresh: true
type: query
rows:
- title: Description
height: 150px
panels:
- title: Description
content: |
Inap Nodepool Status
==========================
This dashboard monitors the status of the nodepool environment for Inap.
**This dashboard is managed by [Grafyaml](http://docs.openstack.org/infra/system-config/grafyaml.html).**
If you would like to make changes to this dashboard, please see the template in the `grafana` directory in
[project-config](https://git.openstack.org/cgit/openstack-infra/project-config/tree/grafana/nodepool.template).
type: text
- title: Nodes
showTitle: true
height: 150px
panels:
- title: Building
span: 3
sparkline:
full: true
show: true
targets:
- target: sumSeries(stats.gauges.nodepool.provider.$region.nodes.building)
type: singlestat
valueName: current
- title: Ready
span: 3
sparkline:
full: true
show: true
targets:
- target: sumSeries(stats.gauges.nodepool.provider.$region.nodes.ready)
type: singlestat
valueName: current
- title: In Use
span: 3
sparkline:
full: true
show: true
targets:
- target: sumSeries(stats.gauges.nodepool.provider.$region.nodes.in-use)
type: singlestat
valueName: current
- title: Deleting
span: 3
sparkline:
full: true
show: true
targets:
- target: sumSeries(stats.gauges.nodepool.provider.$region.nodes.deleting)
type: singlestat
valueName: current
- title: Test Nodes
height: 400px
panels:
- title: Test Node History - $region
type: graph
span: 12
stack: true
repeat: region
minSpan: 4
tooltip:
value_type: individual
yaxes:
- label: "nodes"
- show: false
targets:
- target: alias(sumSeries(stats.gauges.nodepool.provider.$region.nodes.building), 'Building')
- target: alias(sumSeries(stats.gauges.nodepool.provider.$region.nodes.ready), 'Available')
- target: alias(sumSeries(stats.gauges.nodepool.provider.$region.nodes.in-use), 'In Use')
- target: alias(sumSeries(stats.gauges.nodepool.provider.$region.nodes.used), 'Used')
- target: alias(sumSeries(stats.gauges.nodepool.provider.$region.nodes.deleting), 'Deleting')
- target: alias(sumSeries(stats.gauges.nodepool.provider.$region.max_servers), 'Max')
seriesOverrides:
- alias: Max
stack: False
- title: Node Launches
showTitle: true
height: 250px
panels:
- title: Ready Node Launch Attempts
type: graph
span: 4
lines: false
bars: true
nullPointMode: null as zero
yaxes:
- label: "events / min"
- show: false
targets:
- target: aliasSub(summarize(stats_counts.nodepool.launch.provider.$region.ready, '1m'), '.*stats_counts.nodepool.launch.provider.(.*).ready.*', '\1')
- title: Time to Ready
type: graph
span: 4
lines: false
bars: true
nullPointMode: null as zero
yaxes:
- label: "time"
format: ms
- show: false
targets:
- target: aliasByNode(stats.timers.nodepool.launch.provider.$region.ready.mean, 5)
- title: Error Node Launch Attempts
type: graph
span: 4
lines: false
bars: true
nullPointMode: null as zero
yaxes:
- label: "events / min"
- show: false
targets:
- target: alias(smartSummarize(sumSeries(stats_counts.nodepool.launch.provider.$region.error.*), '1m'), "All Errors")
- title: API Operations
showTitle: true
height: 250px
panels:
- title: Create Server
type: graph
lines: true
nullPointMode: connected
span: 4
yaxes:
- format: ms
label: Time
- show: false
targets:
- target: aliasByNode(stats.timers.nodepool.task.$region.ComputePostServers.mean, 4)
- title: Get Server
type: graph
span: 4
lines: true
nullPointMode: connected
span: 4
yaxes:
- format: ms
label: Time
- show: false
targets:
- target: aliasByNode(stats.timers.nodepool.task.$region.ComputePostServers.mean, 4)
- title: Delete Server
type: graph
span: 4
lines: true
nullPointMode: connected
span: 4
yaxes:
- format: ms
label: Time
- show: false
targets:
- target: aliasByNode(stats.timers.nodepool.task.$region.ComputePostServers.mean, 4)
- title: List Servers
type: graph
lines: true
nullPointMode: connected
span: 4
yaxes:
- format: ms
label: Time
- show: false
targets:
- target: aliasByNode(stats.timers.nodepool.task.$region.ComputePostServers.mean, 4)
- title: Get Limits
type: graph
lines: true
nullPointMode: connected
span: 4
yaxes:
- format: ms
label: Time
- show: false
targets:
- target: aliasByNode(stats.timers.nodepool.task.$region.ComputePostServers.mean, 4)