Datasource Examples

Queries relating to datasources, datasource coverage, and which devices have a datasource applied.

See also:

• devices.md for device filtering patterns
• collectors.md for collector-side datasource queries

• Find devices that don’t have a datasource applied
• Count instances and datapoints on a device
  • Instance count
  • Datapoint count
• meta

Find devices that don’t have a datasource applied

Use AssociatedDeviceListByDataSourceId to get devices that DO have the datasource, then compare client-side against your target device set to find those that don’t.

Use an exact name filter (name:) on DatasourceList to avoid retrieving all datasources — there are typically more than 1000.

# Step 1: get the datasource ID
ds_id=$(elm DatasourceList -s0 -f id -F name:NTPv4 | jq -r '.DatasourceList[].id')

# Step 2: get IDs of devices that have it
ntp_ids=$(elm AssociatedDeviceListByDataSourceId --id $ds_id -f id | \
  jq '[.AssociatedDeviceListByDataSourceId[].id]')

# Step 3: find Linux devices that don't have it
elm DeviceList -s0 -f id,displayName \
  -F systemProperties.name:system.sysinfo,systemProperties.value~Linux | \
  jq -r --argjson ntp_ids "$ntp_ids" \
    '.DeviceList[] | select(.id as $id | $ntp_ids | contains([$id]) | not) | .displayName' | sort

The device filter in step 3 can be swapped for any group or property filter — system.sysinfo~Linux is the reliable way to scope to Linux devices rather than relying on group membership, which may include non-Linux devices.

To also show host status and collector — useful for triaging whether missing coverage is because the device is dead or just never had the datasource applied:

# Step 1: get the datasource ID
ds_id=$(elm DatasourceList -s0 -f id -F name:NTPv4 | jq -r '.DatasourceList[].id')

# Step 2: get IDs of devices that have it
ntp_ids=$(elm AssociatedDeviceListByDataSourceId --id $ds_id -f id | \
  jq '[.AssociatedDeviceListByDataSourceId[].id]')

# Step 3: find Linux devices that don't have it, with status and collector
elm DeviceList -s0 -f id,displayName,hostStatus,collectorDescription \
  -F systemProperties.name:system.sysinfo,systemProperties.value~Linux | \
  jq -r --argjson ntp_ids "$ntp_ids" \
    '.DeviceList[] | select(.id as $id | $ntp_ids | contains([$id]) | not) | [.displayName, .hostStatus, .collectorDescription] | @tsv' | \
  sort | column -t -s$'\t'

Devices with hostStatus: normal are the real gaps — dead or dead-collector devices won’t collect NTP data regardless of whether the datasource is applied.

Count instances and datapoints on a device

These are different things:

• Instance — a monitored object, e.g. the filesystem /var on a device
• Datapoint — an individual metric within an instance, e.g. SpaceUsed, SpaceUsedPercent

Instance count

One API call using -C:

elm DeviceInstanceList --id <deviceId> -C

Datapoint count

No single endpoint returns total datapoints. This pipeline groups instances by datasource, fetches each unique datasource definition once, and multiplies datapoints × instance count:

elm DeviceInstanceList --id <deviceId> -s0 \
  | jq -r '.DeviceInstanceList | group_by(.dataSourceId)[] | "\(.[0].dataSourceId) \(length)"' \
  | while read ds_id inst_count; do
      dp_count=$(elm DatasourceById --id $ds_id -f dataPoints \
        | jq '.DatasourceById[0].dataPoints | length')
      echo $((dp_count * inst_count))
    done \
  | awk '{sum += $1} END {print sum}'

API calls: 1 (DeviceInstanceList) + 1 per unique datasource on the device — typically much fewer than the total instance count.

Note: this counts all configured datapoints. It does not verify that every instance has active data flowing.

meta

Update the ToC on this page by running the following:

gh-md-toc --insert --no-backup --hide-footer --skip-header examples/datasources.md