"Telegraf host metrics" mixes hosts on some graphs even when host filtering is applied

[Alvaro Uria]

Several graphs, such as Network cards drops/errors list metric(s) for __all hosts__ instead of applying '''host =~ "$host"''' filtering.

In addition to the above, legends not always list "{{host}} {{interface}}", which becomes handy when more than a single node is selected.

"Telegraf host metrics" template needs to be fixed.

Possible testing:
1) Deploy telegraf in a test env from scratch and verify all metrics are related to the selected host
2) Upgrade telegraf charm in an env using the previous version and see if the "[Juju] xxxx" dashboards get updated (specially, the "[Juju] Telegraf host metrics" one.

[Xav Paice (xavpaice)]

Confirmed using current charms, the 'Network Throughput' graph shows stats for all hosts when only one is selected. I was unable to verify the drops/errors since my test VMs had no drops/errors to report.

[Andrea Ieri (aieri)]

The 'Network Throughput' is additionally incorrectly using the same query net_bytes_recv for both egress and ingress.

Queries should rather be:
A: irate(net_bytes_recv{host =~ "$host"}[1m])
B: irate(net_bytes_sent{host =~ "$host"}[1m])

[Xav Paice (xavpaice)]

dashboard template for this is `"expr": "irate(net_bytes_recv{host=~\"$host\"}[1m])"` in the current Telegraf master, however there's a bug which is preventing changes to the dashboard being propagated to the Grafana instance (LP:#1919937).

[Xav Paice (xavpaice)]

Added charm-telegraf as the dashboard for Telegraf metrics now lives in the Telegraf charm, not Grafana.