Generic monitoring

[srvstrnull]

I understand the concept, is it also possible to monitor/consume metrics/logs of bare-metal/virtual-machines/network appliances (snmp), that are not part of for eg a Kubernetes cluster?

[TayfurYldz]

Yes. HyperDX is not limited to telemetry produced inside Kubernetes; the storage/query side consumes OpenTelemetry data, so bare-metal and VM workloads can send logs/metrics/traces as long as you get them into an OpenTelemetry Collector that exports to HyperDX.

For hosts, the usual collector-side building blocks are the OpenTelemetry Collector's host metrics receiver (CPU, memory, filesystem, network, etc.) plus a file/log receiver where needed. Those can run on the VM/bare-metal host or on a reachable collector tier.

SNMP is slightly different: HyperDX itself is not an SNMP poller. For network appliances you need an SNMP-capable collector/receiver or another SNMP exporter that converts the polled values into OTLP/OpenTelemetry metrics, then send those metrics to the HyperDX ingestion path.

The current HyperDX architecture in CONTRIBUTING.md makes the boundary fairly clear: its otel-collector receives OpenTelemetry data and forwards it to ClickHouse. The source of that telemetry does not have to be a Kubernetes pod.

So a mixed environment can look roughly like:

VM / bare metal -> OTel Collector ---------\
                                           -> HyperDX ingestion -> ClickHouse
network device -> SNMP poller -> OTLP -----/

The main thing I would avoid is expecting HyperDX to directly discover/poll arbitrary SNMP devices. Keep protocol-specific collection at the collector/exporter layer and use OTLP as the boundary into HyperDX.