vcenterreceiver : Get the total OS uptime in the 5m interval

[manjunath-batakurki]

Hi Team,

I'm looking for the ways to get the total OS uptime in the given interval(5 minutes in my case). In my case, the scraping happens every 5 minz and based on the vCenter metric, i'm able to get the OS Uptime that is returned by the vCenter and return to user.

The challenge is that, the VM could be down/up multiple times between the scrapping interval, for example, In a 5min scrapping interval duration, if the VM was down at the second minute, and was again up at the 4th minute, the value returned would be just 60 seconds uptime(actually it was up for 120 seconds), is there a way to get the total uptime ?

below is the metric given by vCenter as part of Performance Counter.
`
metrics:

 vcenter.vm.sys.osUptime.latest:
   enabled: true
   description: Total time elapsed, in seconds, since last operating system boot-up
   unit: second
   sum:
     monotonic: false
     value_type: int
     aggregation: cumulative
   attributes: [vm_name]

is it possible to get the total uptime in the scrapping interval ?

cat /etc/otel-collector-config.yaml

receivers:
  vcenter:
    endpoint: "https://<ip>/"
    username: '<username>
    password: '<Passwword>'
    collection_interval: 5m
    tls:
      insecure: true
      insecure_skip_verify: true
 
exporters:
  logging:
    loglevel: debug
  prometheusremotewrite:
    endpoint: "https://endpoint"
    headers:
      Authorization: "Bearer <token>"
 
service:
  pipelines:
    metrics:
      receivers: [vcenter]
      exporters: [logging, prometheusremotewrite]

[manjunath-batakurki]

@schmikei Do you have any inputs here ?

[schmikei]

Hmm so the collector will always collect at an interval, you can try reducing the interval, but that will potentially cause performance issues for either the collector. There's no event based scraping for metrics so inherently you'll be estimating the "total uptime" value irregardless since between when we make the API call and the next scrape, the VM could have powered on/off a couple times and from the collector's perspective it wouldn't necessarily know.

Depending on which platform (like a prometheus exporter or Google Cloud Monitoring) is ingesting these metrics, they may have query tools to aggregate only positive deltas in data points (albeit I can imagine that's could be one complex query depending on the platform). But still its going to be an estimate in essence as we'd miss the data points in between a scrape interval.

There may also be a processor based approach that potentially uses the transformprocessor, but would need to look into it more.

I was able to grab a value for the metric with a simple addition to the current receiver however whether or not this solves your use case or is data model correct may need more exploration.

            {
              "name": "vcenter.vm.sys.uptime",
              "description": "Total time elapsed, in seconds, since last operating system boot-up.",
              "sum": {
                "dataPoints": [
                  {
                    "startTimeUnixNano": "1662062317306012000",
                    "timeUnixNano": "1662062420000000000",
                    "asInt": "12191805"
                  }
                ],
                "aggregationTemporality": "AGGREGATION_TEMPORALITY_CUMULATIVE"
              }
            }

I'd usually defer to someone who knows the data model a tad more than I do for cases like these, so I'll ask around to get a couple other perspectives!