Varnish Cache¶

The varnish probe monitors Varnish Cache via varnishstat -j -1, reporting cache hit/miss rates, client request throughput, backend connections, thread lifecycle, session counts, object counts and memory allocation.

Quick start¶

# probes.d/10-varnish.yaml — each file under probes.d/ is a YAML array of probes
- name: varnish
  type: varnish

No parameters are required for a single-instance Varnish installation. The probe runs varnishstat as the agent user — ensure it has access to the Varnish shared memory file.

Parameters¶

Parameter	Must set	Default	Description
`varnishstat_path`	No	`varnishstat`	Path of the varnishstat binary when it is not on the PATH. Example: `/usr/bin/varnishstat`
`instance_name`	No	-	Varnish instance name passed as -n; needed when several instances run on the host
`interval`	No	`60`	Seconds between collections

Metrics¶

Metric	Unit	Description
`senhub.varnish.up`	1	1 when `varnishstat` completed successfully
`varnish.cache.operations`	{operation}	Cache lookups by result (hit/miss/hitpass), tagged with `result`
`varnish.client.requests.received`	{request}	Client requests received
`varnish.backend.connections.fail`	{connection}	Failed backend connection attempts
`varnish.thread.operations`	#	Thread lifecycle events, tagged with `operation` (created, destroyed, failed)
`varnish.session.connections`	#	Client sessions accepted
`varnish.session.dropped`	{session}	Sessions dropped due to overflow
`varnish.objects.stored`	{object}	Objects currently stored in cache
`varnish.memory.allocated`	By	Memory allocated for cache storage

Operational notes¶

The varnishstat command reads from the Varnish shared memory log; the user running the agent must be in the varnish group or run as root.
instance_name maps to varnishstat -n <name> and is needed when you run multiple Varnish instances (different working directories via varnishd -n).
Metric names align with the OpenTelemetry Collector contrib varnishreceiver where equivalents exist.

Metric reference¶

Every metric this probe can emit. Metric is the OpenTelemetry name the OTLP, Prometheus and Zabbix outputs derive theirs from. Name is what a Nagios check and the API metrics= filter match. PRTG channel is the label PRTG shows, placeholders filled from the series' tags.

Metric	Name	PRTG channel	Unit	Description
`senhub.varnish.up`	`senhub.varnish.up`	Varnish Up	#	1 when varnishstat completed successfully, 0 otherwise
`varnish.cache.operations`	`varnish.cache.operations`	Varnish Cache	#	Number of cache lookup results by type (hit, miss, hitpass)
`varnish.client.requests.received`	`varnish.client.requests.received`	Varnish Client Requests	#	Total client requests received
`varnish.backend.connections.success`	`varnish.backend.connections.success`	Varnish Backend Connections Success	#	Backend connections successfully established
`varnish.backend.connections.fail`	`varnish.backend.connections.fail`	Varnish Backend Connections Failed	#	Backend connection attempts that failed
`varnish.backend.connections.reused`	`varnish.backend.connections.reused`	Varnish Backend Connections Reused	#	Backend connections reused from keep-alive pool
`varnish.thread.operations`	`varnish.thread.operations`	Varnish Threads	#	Thread lifecycle events by operation (created, destroyed, failed)
`varnish.session.connections`	`varnish.session.connections`	Varnish Session Connections	#	Accepted client sessions
`varnish.session.dropped`	`varnish.session.dropped`	Varnish Session Dropped	#	Client sessions dropped because the session queue was full
`varnish.objects.stored`	`varnish.objects.stored`	Varnish Objects Stored	#	Number of HTTP objects currently stored in the cache
`varnish.memory.allocated`	`varnish.memory.allocated`	Varnish Memory Allocated	bytes	Total bytes currently allocated across all storage allocators (SMA/SMF)