VERSION. 1.7.8
DESCRIPTION. pmacct is a small set of multi-purpose passive network monitoring tools. It can account, classify, aggregate, replicate and export forwarding-plane data, ie. IPv4 and IPv6 traffic; collect and correlate control-plane data via BGP and BMP; collect and correlate RPKI data; collect infrastructure data via Streaming Telemetry. Each component works both as a standalone daemon and as a thread of execution for correlation purposes (ie. enrich NetFlow with BGP data). A pluggable architecture allows to store collected forwarding-plane data into memory tables, RDBMS (MySQL, PostgreSQL, SQLite), noSQL databases (MongoDB, BerkeleyDB), AMQP (RabbitMQ) and Kafka message exchanges and flat-files. pmacct offers customizable historical data breakdown, data enrichments like BGP and IGP correlation and GeoIP lookups, filtering, tagging and triggers. Libpcap, Linux Netlink/NFLOG, sFlow v2/v4/v5, NetFlow v5/v8/v9 and IPFIX are all supported as inputs for forwarding-plane data. Replication of incoming NetFlow, IPFIX and sFlow datagrams is also available. Collected data can be easily exported (ie. via Kafka) to modern databases like ElasticSearch, Apache Druid and ClickHouse and (ie. via flat-files) to classic tools Cacti, RRDtool and MRTG, etc. Control-plane and infrastructure data, collected via BGP, BMP and Streaming Telemetry, can be all logged real-time or dumped at regular time intervals to AMQP (RabbitMQ) and Kafka message exchanges and flat-files. HOMEPAGE. http://www.pmacct.net/ DOWNLOAD. http://www.pmacct.net/pmacct-1.7.8.tar.gz CHANGELOG. + Introduced support for eBPF for all daemons: if SO_REUSEPORT is supported by the OS and eBPF support is compiled in, this allows to load a custom load-balancer. To load-share, daemons have to be part of the same cluster_name and each be configured with a distinct cluster_id. + Introduced support for listening on VRF interfaces on Linux for all daemons. The feature can be enabled via nfacctd_interface, bgp_daemon_interface and equivalent knobs. Many thanks to Marcel Menzel ( @WRMSRwasTaken ) for this contribution. + pre_tag_map: introduced limited tagging / labelling support for BGP (pmbgpd), BMP (pmbmpd), Streaming Telemetry (pmtelemetryd) daemons. ip, set_tag, set_label keys being currently supported. + pre_tag_map: defined a new pre_tag_label_encode_as_map config knob to encode the output 'label' value as a map for JSON and Apache Avro encodings, ie. in JSON "label": { "key1": "value1", "key2": "value2" }. For keys and values to be correctly mapped, the '%' delimiter is used when composing a pre_tag_map, ie. "set_label=key1%value1,key2%value2 ip=0.0.0.0/0". Thanks to Salvatore Cuzzilla ( @scuzzilla ) for this contribution. + pre_tag_map: introduced support for IP prefixes for src_net and dst_net keys for indexed maps (maps_index set to true). Indexing being an hash map, this feature currently tests data against all defined IP prefix lenghts in the map for a match (first defined matching prefix wins). + pre_tag_map: introduced two new 'is_nsel', 'is_nel' keys to check for the presence of firewallEvent field (233) and natEvent field (230) in NetFlow/IPFIX respectively in order to infer whether data is NSEL / NEL. If set to 'true' this does match NSEL / NEL data, if set to 'false' it does match non NSEL / NEL data respectively. + Introduced a new mpls_label_stack primitive, encoded as a string and includes a comma-separated list of integers (label values). Thanks to Salvatore Cuzzilla ( @scuzzilla ) for this contribution. + Introduced a new fw_event primitive, to support NetFlow v9/ IPFIX firewallEvent 233 Information Element. + Introduced a new tunnel_tcp_flags primitive for pmacctd and sfacctd to record TCP flags for the inner layer of a tunneled technology (ie. VXLAN). Also tunnel_dst_port decoding was fixed for sfacctd. + Introduced support for in/out VLAN support for sfacctd. To be savy, 'in_vlan' and 'vlan' were muxed onto the same primitive depending on the daemon being used. Thanks to Jim Westfall ( @jwestfall69 ) for this contribution. + Introduced a new mpls_label_stack_encode_as_array config knob to encode the MPLS label stack as an array for JSON and Apache Avro encodings, ie. in JSON "mpls_label_stack": [ "0-label0", "1-label1", "2-label2", "3-label3", "4-label4", "5-label5" ] and in Avro "name": "mpls_label_stack", "type": { "type": "array", "items": { "type": "string" } }. Thanks to Salvatore Cuzzilla ( @scuzzilla ) for this contribution. + Introduced a new tcpflags_encode_as_array config knob to encode TCP flags as an array for JSON and Apache Avro, ie. in JSON "tcp_flags": [ "URG", "ACK", "PSH", "RST", "SYN", "FIN" ] and in Avro "name": "tcp_flags", "type": { "type": "array", "items": { "type": "string" } }. Thanks to Salvatore Cuzzilla ( @scuzzilla ) for this contribution. + Introduced a new fwd_status_encode_as_string config knob to encode the 'fwd_status' primitive in human-readable format like described by RFC-7270 Section 4.12 when JSON or Avro formats are selected for output. Thanks to Salvatore Cuzzilla ( @scuzzilla ) for this contribution. + Introduced a new protos_file to define a list of (known/ interesting/meaningful) IP protocols. Both protocol names, ie. "tcp", and protocol numbers, ie. 1 (for icmp), are accepted. IANA reserved protocol value 255 is used to bucket as 'others' those IP protocols not matching the ones defined in the list. + Introduced a new tos_file to define a list of (meaningful) IP ToS values; if tos_encode_as_dscp is set to true then DSCP values are expected as part of the file. The directive uses value 255 to bucket as 'others' those ToS/DSCP values not matching the ones defined in the list. + A new tos_encode_as_dscp config knob makes pmacct to honour only the 6 bits used by DSCP and report only on those. + BGP, BMP, Streaming Telemetry daemons: introduced a new dump_time_slots config knob to spread the load deriving by dumps over the configured refresh time interval. The interval is divided into time slots and nodes are assigned to such slots. The slot for each node is determined using its IP address. Thanks to Raphael Barazzutti ( @rbarazzutti ) for this contribution. + BGP, BMP daemons: End-of-RIB messages are now being exposed in the output feed in order to facilitate tracking their arrival (or not!). + pmtelemetryd: aligned daemon to the latest Unyte UDP-Notif API (0.6.1) and related standardization draft-ietf-netconf-udp-notif + RPKI daemon: added case for input "asn" value being integer (ie. "asn" : 2914) on top of the string case (ie. "asn" : "AS2914"). + Kafka, amqp plugins: introduced a new writer_id_string config knob to allow to customize the the "writer_id" field value. A few variables are supported along with static text definitions. + Added a new aggregate_unknown_etype config knob to account also frames with EtherTypes for which there is no decoding support and allow to aggregate them by the available Ethernet L2 fields (ie. 'src_mac', 'dst_mac', 'vlan', 'cos', 'etype'). Thanks to @singularsyntax for this contribution. + Added a new bgp_daemon_add_path_ignore config knob to ignore (do not advertise back) the ADD-PATH capability advertised by remote BGP peers. + nfacctd, sfacctd: extended the possibility to run daemons from a user with non root privileges to these daemons. + nfacctd: if Information Element 90 (MPLS VPN RD) is present in NetFlow v9/IPFIX, make it available for BGP/BMP correlation. + pmacctd, sfacctd: introduced basic support for QinQ, 802.1AD. + [print|kafka|amqp]_preprocess: added suppport for 'maxp', 'maxb' and 'maxf' keys when preprocessing aggregates of non- SQL plugins. Thanks to Andrew R. Lake ( @arlake228 ) for this contribution. + nDPI: newer versions of the library (ie. >= 4.0) bring changes to the API. pmacct is now aligned to compile against these. At the same time support for nDPI 3.x was dropped. ! fix, plugin_common.[ch]: when stitching feature was enabled, ie. nfacctd_stitching, timestamp_min was never reset. Also both timestamp_min and timestamp_max were clamped to sec granularity. ! fix, BGP, BMP daemons: added a tmp_bgp_daemon_origin_type_int to print out BGP "origin" field as int (legacy behaviour) instead of string (current behaviour). In a future major release the legacy behaviour will be dropped. ! fix, BGP, BMP daemons: MPLS labels are now encoded in both JSON and Apache Avro as 'mpls_label' instead of 'label'. This is to align behaviour with pre_tag_map where 'label' has a different semantic. ! fix, BGP, BMP daemons: resolved memory leak when encoding log messaging (logmsg) in Avro format with Schema Registry support. ! fix, BGP daemon: improved handling of ADD-PATH capability, making it per-AF (as it is supposed to be) and not global. ! fix, BMP daemon: now checking that ADD-PATH capability is enabled at both ends of the monitored session (check both BGP OPEN in a Peer Up message) in order to infer that the capability exchange was successful. Also some heuristics were added to conciliate BGP Open vs BGP Update 4-bytes ASN reality. ! fix, nfacctd: improved parsing of NetFlow v9 Options data particularly when multiple IEs are packed as part of a flowset. ! fix, nfacctd: corrected parsing of Information Element 351 (layer2SegmentId). ! fix, pmacctd: improved processing of pcap_interfaces_map for cases where the same interface is present multiple times (maybe with different directions). Also, if the map is empty then bail out at startup. ! fix, pmacctd: SEGV when ICMP/ICMPv6 traffic was processed and 'flows' primitive was enabled. ! fix, pmacctd: sampling_rate primitive value was not reported correctly when 'sampling_rate' config directive was specified. ! fix, pmbgpd, pmpmbd, pmtelemtryd: changed SIGCHLD handler to prevent zombification of last spawned data dump writer. ! fix, Kafka plugin: moved the schema registration from the dump writer to the plugin process in order to register the schemas only once at plugin startup and not on every start of a writer process. Thanks to Uwe Storbeck ( @ustorbeck ) for this contribution. ! fix, Kafka plugin: a check for kafka_partition was missing, leading the plugin to always use the default partitioner instead of sending data to the configured fixed partition. Thanks to Martin Pels ( @rodecker ) for this contribution. ! fix, nfprobe plugin: BGP data enrichment was not working due to a mistakenly moved pointer. ! fix, sfprobe plugin: AS-PATH was being populated even when null; added a check to see if the destination AS is not zero in order to put the destination AS into the AS-PATH for sFlow packets. Thanks to Marcel Menzel ( @WRMSRwasTaken ) for this contribution. ! fix, networks_file: remove_dupes() was making partial commits of valid rows hence creating data inconsistencies. ! fix, pre_tag_map: resolved a potential string overflow that was being triggered in pretag_append_label() when data would be assigned more than one single label. Also now allow ',' chars in set_label. ! fix, maps_index: uninitialized var could cause SEGV in case no results are found in the map index. Also introduced support for catch-all rules, ie. "set_label=unknown". ! fix, maps_index: optimized the case of no 'ip' key specified (for nfacctd and sfacctd): when indexing is enabled, prevent recirculation from happening, ie. test v4 first then v6, since the 'ip' key is not going to be part of the hash serializer. ! fix, pretag.c: allow to allocate maps greater than 2GB in size. Also several optimizations were carried out yelding to a better memory utilization for allocated maps along with improved times to resolve JEQs. ! fix, pre_tag_label_filter: optimized and improved runtime evaluation part of this feature, avoiding a costly strdup() and returning immediately on certain basic mismatch conditions. ! fix, kafka_common.[ch]: a new p_kafka_produce_data_and_free() is invoked to optimize memory allocations and releases. ! fix, plugin_cmn_avro.c: when a schema registry is being defined, ie. kafka_avro_schema_registry, the logic to generate the schema name has been changed: use topic plus record name as the schema name, use underscore as separator within the record name, stop adding a "-value" suffix. Thanks to Uwe Storbeck ( @ustorbeck ) for this contribution. ! fix, util.c: roundoff_time() to reason always with the locally configured time, like for the rest of functional (as in non-data) timestamps, ie. refresh time, deadline, etc. ! fix, log.c: when log messages are longer than message buffer, the message gets cut off. As the trailing newline also gets cut off the message will be concatenated with the following message which makes the log hard to read. Thanks to Uwe Storbeck ( @ustorbeck ) for this contribution. - Completed the retirement of legacy packet classification based on home-grown code (Shared Objects) and the L7 layer project. - Removed the mpls_stck_depth primitive due to the introduction of the mpls_label_stack primitive. NOTES. See UPGRADE file. Cheers, Paolo _______________________________________________ pmacct-discussion mailing list http://www.pmacct.net/#mailinglists