Skip to main content

Plugin Health Charts

The netflow plugin publishes its own operational charts under the netdata.netflow.* chart context. These appear on the standard Netdata charts page (alongside system metrics like CPU and memory), not inside the Network Flows view. They are how you monitor the plugin itself: is it receiving data, are templates flowing, is memory growing, is disk being written.

This is also where you look first when something seems wrong — long before opening the Network Flows view.

Default production-health charts update every 1 second. Optional memory diagnostics charts update on charts.memory_diagnostics.interval, rounded up to whole seconds, when enabled.

The charts​

The dashboard shows the charts under Netdata > Plugins > netflow, in six groups that follow the data path: input, decoding, cisco nsel, storage, state, and memory. The table lists them in that order.

GroupChartTypeWhat it shows
inputnetdata.netflow.input_packetsline, packets/sUDP datagrams received, empty datagrams, and Linux socket-buffer drops
inputnetdata.netflow.input_bytesline, bytes/sUDP byte rate
inputnetdata.netflow.protocol_packetsstacked, packets/sSuccessfully identified v5, v7, v9, IPFIX, and sFlow packets
decodingnetdata.netflow.flow_setsstacked, sets/sNetFlow v9 and IPFIX Data, Options Data, Template, missing-template, and ignored Sets
decodingnetdata.netflow.templatesstacked, templates/sData and Options Template definitions received over v9 and IPFIX
decodingnetdata.netflow.flow_recordsstacked, records/sv5, v7, v9, and IPFIX Data Records decoded
decodingnetdata.netflow.options_recordsstacked, records/sv9/IPFIX Options Records and sampling configuration sent in Data Records
decodingnetdata.netflow.sflow_samplesstacked, samples/ssFlow flow, counter, discarded-packet, real-time, and unknown samples
decodingnetdata.netflow.decoder_exceptionsline, events/sReceive, parse, template, disabled-protocol, eviction, counter, decapsulation, and unsupported-data exceptions
decodingnetdata.netflow.decoder_scopesline, scopesDistinct (exporter, observation domain) scopes the decoder tracks
decodingnetdata.netflow.flow_rowsline, rows/sDecoded rows and their final filtered, journaled, or write-failed outcomes
cisco nselnetdata.netflow.nsel_eventsstacked, records/sCisco NSEL update, create, teardown, deny, unsupported, and malformed records
cisco nselnetdata.netflow.nsel_rowsstacked, rows/sForward and reverse rows emitted from NSEL updates
cisco nselnetdata.netflow.nsel_exceptionsline, events/sCounterless updates, partial counter directions, and suppressed zero responders
storagenetdata.netflow.raw_journal_opsline, ops/sRaw journal: writes, sync calls, errors
storagenetdata.netflow.raw_journal_bytesline, bytes/sRaw journal logical bytes written
storagenetdata.netflow.materialized_tier_opsline, ops/sRollup tiers: rows produced per tier, flushes, errors
storagenetdata.netflow.materialized_tier_bytesstacked, bytes/sRollup tier byte rate, broken down by tier
storagenetdata.netflow.tier_commit_ageline, secondsTime since each tier worker completed a commit
storagenetdata.netflow.tier_commit_durationline, microsecondsDuration of each tier's most recent commit
storagenetdata.netflow.tier_commit_batchesline, batches/sCompleted commit batches per tier
storagenetdata.netflow.tier_commit_stretchedline, events/sTier commits that exceeded their planned window
storagenetdata.netflow.journal_io_opsline, ops/sDecoder-state persistence and facet-index operations/errors
storagenetdata.netflow.journal_io_bytesline, bytes/sDecoder-state persist byte rate
statenetdata.netflow.open_tiersstacked, rowsRows currently open per tier
statenetdata.netflow.tier_index_entriesline, entriesTier-index hours and rollup-flow entries
statenetdata.netflow.facet_valuesline, valuesPublished facet value cardinality
statenetdata.netflow.facet_fieldsline, fieldsPopulated and autocomplete-backed facet fields
memorynetdata.netflow.memory_resident_bytesline, bytesOptional diagnostics: process RSS, peak RSS, breakdown
memorynetdata.netflow.memory_resident_mapping_bytesstacked, bytesOptional diagnostics: RSS broken down by mapping type
memorynetdata.netflow.memory_allocator_bytesline, bytesOptional diagnostics: allocator-internal stats
memorynetdata.netflow.memory_accounted_bytesstacked, bytesOptional diagnostics: RSS attributed to known components, plus unaccounted
memorynetdata.netflow.memory_tier_index_bytesstacked, bytesOptional diagnostics: tier-index memory drilldown

Reading the most useful charts​

Input and protocol packets​

netdata.netflow.input_packets describes the UDP socket only:

  • udp_received — every datagram pulled from a listener socket, including empty datagrams.
  • empty — received zero-length datagrams. This is a subset of udp_received.
  • kernel_dropped — datagrams Linux dropped because a specific listener socket's receive buffer was full. The plugin reads the kernel counter once per second by exact socket identity. It is zero on platforms that do not expose this counter.

netdata.netflow.protocol_packets reports successfully identified v5, v7, v9, IPFIX, and sFlow packets separately. These counters still increase when that protocol is disabled for row production, so disabled_protocol_packets can explain why recognized traffic did not reach storage.

Sets, templates, records, and samples​

Do not compare these rates as though they were the same thing:

  • One UDP datagram can contain multiple v9/IPFIX Sets.
  • One Set can contain multiple template definitions or Data Records.
  • One sFlow datagram can contain multiple samples.
  • One Data Record can produce zero, one, or two stored rows.

Use netdata.netflow.flow_sets to verify v9/IPFIX control traffic. *_missing_template means a Data Set could not be decoded because its template was unavailable. v9_ignored is a reserved Set ID; ipfix_ignored covers template withdrawals and reserved Set IDs.

Use netdata.netflow.templates to verify that exporters refresh both Data and Options Templates. Use netdata.netflow.flow_records for decoded Data Records, netdata.netflow.options_records for sampling/options traffic, and netdata.netflow.sflow_samples to distinguish flow-bearing sFlow samples from valid counter-only traffic. A Set first received without a template is counted as missing; if the parser later replays it after learning the template, its decoded records are counted then.

netdata.netflow.flow_rows​

This chart follows records after decoding. Its cumulative counters obey:

decoded = classifier_filtered + journaled + write_failed

If records are visible but rows are not:

  • classifier_filtered means configured enrichment/classifier policy rejected them.
  • write_failed means the journal could not store them.
  • A gap before decoded is explained by the protocol-specific charts and netdata.netflow.decoder_exceptions.

Cisco NSEL charts​

NSEL records and traffic rows are intentionally not one-to-one:

  • netdata.netflow.nsel_events counts every recognized event record by type.
  • netdata.netflow.nsel_rows counts forward and reverse traffic rows emitted from event-5 updates.
  • netdata.netflow.nsel_exceptions explains update records that produced fewer rows because counters were missing, partial, or zero.

Create, teardown, and deny events are visible in netdata.netflow.nsel_events but are not stored as traffic rows.

netdata.netflow.decoder_exceptions​

All dimensions are exceptional events, but they occur at different stages:

  • udp_receive_errors, udp_socket_setup_errors, and parse_errors identify socket/parser failures.
  • missing_template_sets identifies v9/IPFIX Sets received without an available template.
  • disabled_protocol_packets identifies valid packets intentionally excluded by configuration.
  • parser_source_evictions identifies exporter parser state removed at the configured source limit.
  • partial_counter_records, decapsulation_failed_records, unsupported_data_sets, and ipfix_zero_reverse_records explain record-to-row differences.

Because these dimensions count different objects, use them to identify causes; do not add them together or divide all of them by one common denominator.

netdata.netflow.decoder_scopes​

Cardinality of decoder state. Reports how many distinct (exporter, observation domain) template caches the plugin currently holds. Watch for unbounded growth — an exporter that frequently rotates template IDs (rare but real) will inflate this without bound.

netdata.netflow.facet_*, netdata.netflow.tier_index_entries, and netdata.netflow.open_tiers​

Default lightweight production signals. Use them to see whether memory pressure is likely coming from growing facet vocabulary, rollup-index cardinality, or open rollup rows without enabling expensive byte-level diagnostics.

netdata.netflow.materialized_tier_*​

Show the rollup pipeline working. *_rows should track ingest. flushes should tick steadily; if it stops, tiering is stalled.

netdata.netflow.memory_resident_bytes and netdata.netflow.memory_accounted_bytes​

These charts exist only when charts.memory_diagnostics.enabled: true is set in netflow.yaml.

If RSS climbs over time:

  • Check netdata.netflow.memory_accounted_bytes to see where it's going.
  • The unaccounted dimension is RSS - sum(known components). A growing unaccounted is your leak signal.
  • tier_indexes and open_tiers are normal sources of growth — they should track ingest rate.
  • geoip_asn and geoip_geo are mmap'd MMDB files. Their size grows as the kernel pages the file in under read pressure.

netdata.netflow.memory_resident_mapping_bytes​

This one breaks RSS down by what's mapped. Useful when you want to attribute "this process is using 800 MB" — heap, journals (per tier), MMDB files, anonymous mappings, etc.

What's NOT in these charts​

These charts do not include:

  • Per-exporter ingest counter. No per-source rate dimension. Decoder-scope cardinality tells you how many sources, not how busy each one is.
  • NIC and network-wide packet loss. kernel_dropped covers only the plugin's Linux UDP listener sockets, not upstream network loss or NIC drops.
  • Template cache hit ratio. Missing-template Sets are counted, but there is no corresponding hit counter.
  • GeoIP staleness signal. No "MMDB last loaded" timestamp or version. The mapping memory dimensions tell you if a database is loaded, not how old it is.
  • Per-tier query latency. These charts cover ingest and storage; query-side performance isn't observable.
  • BioRIS counters. BioRIS routing-state details are not published as chart dimensions.

How to use these charts for diagnosis​

SymptomLook atWhat it means
Network Flows view is emptynetdata.netflow.input_packets, then netdata.netflow.protocol_packetsNo UDP = network/listener problem. UDP without protocol packets = malformed or unsupported traffic; check decoder exceptions.
Sudden drop in flowsnetdata.netflow.protocol_packets, netdata.netflow.flow_records, netdata.netflow.flow_rowsIdentifies whether traffic stopped at the exporter, decoder, classifier, or journal.
Templates failing*_missing_template in netdata.netflow.flow_setsExporter not sending templates often enough, collector has not learned them yet, or exporter identity changed.
sFlow packets but no rowsnetdata.netflow.sflow_samplesCounter-only samples are valid but do not contain endpoint traffic rows.
NSEL records but fewer rowsnetdata.netflow.nsel_events, netdata.netflow.nsel_rows, netdata.netflow.nsel_exceptionsNon-update events are intentionally not traffic; directional counter outcomes explain event-5 rows.
UDP loss under loadkernel_dropped in netdata.netflow.input_packetsLinux listener receive buffer overflow. This does not cover upstream or NIC loss.
Cache growing without bounddecoder_scopes rising over hoursExporter churn or unstable template IDs. Investigate per-router behaviour.
Memory pressurenetdata.netflow.facet_values, netdata.netflow.tier_index_entries, netdata.netflow.open_tiers; optional netdata.netflow.memory_* diagnosticsDefault count charts show where state cardinality grows. If byte diagnostics are enabled and unaccounted grows → unattributed allocation, possibly a leak.
Disk write stallsnetdata.netflow.raw_journal_ops write_errors, sync_errorsDisk full, permission denied, fs error.
Decoder/facet state not persistingnetdata.netflow.journal_io_opsPersist calls should tick periodically; all error dimensions should remain zero.

Where these are NOT shown​

These charts are not in the Network Flows view. Look for them on the standard Netdata charts page, under Netdata > Plugins > netflow. The Network Flows view itself shows traffic data, not plugin health.

What's next​


Do you have any feedback for this page? If so, you can open a new issue on our netdata/learn repository.