- Enable additional dependencies (gprc, json, libcurl, libelf,
libwebsockets, protobuf, snappy, xenstat, yajl)
- Update to v1.28.0 (go.d.plugin 0.26.2)
Release v1.28.0 is a hotfix release to address a deadlock in the
Netdata Agent. If the Agent-Cloud link (ACLK) connection drops and
the Agent fails to queue an on_connect message, it also fails to
properly release a lock in the web server thread.
- Changes in v1.27.0:
The v1.27.0 release of the Netdata Agent brings dramatic
improvements to long-term metrics storage via the database engine,
and new dashboard features like a time & date picker for
visualizing precise timeframes. Two new collectors bring incredible
new value to existing features, including a bit of machine learning
magic.
This release contains 8 new collectors, 1 new notification method
(2 others enhanced), 54 improvements, 41 documentation updates, and
58 bug fixes.
Improvements
* Add labels for Kubernetes pods and containers.
* Add plugin and module health entities.
* Migrate the metadata log to SQLite.
* Add an extent cache to the database engine.
* Added new data query option allow_past.
Netdata Cloud
* Add the ability to query child nodes by their GUID.
* Add child availability messages to the ACLK.
* Add a metric showing how long a query spent in the queue.
* Completely hide the SSO iframe.
Collectors
* Add alarms obsoletion and disable alarms collector by default.
* Add calls for tcp_sendmsg, tcp_retransmit_skb, tcp_cleanup_rcv,
udp_sendmsg, udp_recvmsg functions charts to the eBPF
collector.
* Add two more insignificant warnings to suppress in anomalies
collector.
* Add the number of allocated/stored objects within each storage
to the varnish collector.
* Add a wireless statistics collector.
* Add support for MSE (Massive Storage Engine) to the varnish
collector.
* Remove remove crit from unmatched alarms in the web_log
collector.
* Add GPU key metrics (nvidia_smi collector) to
dashboard_info.js.
* Add allocated space metrics to the oracledb collector.
* Restructure the eBPF collector to improve usability.
* Add an anomaly detection collector.
* Add a Netdata alarms collector.
* Add a configuration option to exclude users with zero memory
allocated to the nvidia_smi collector.
* Add per queue charts to the rabbitmq collector.
* Add support for HBA drives to the hpssa collector.
* Update the cgroups collector default filtering by adding pod
level cgroups.
* Add a Go version of the CouchDB collector (couchdb).
* Add collecting HTTP method per URL pattern (url_pattern option)
to the web_log collector.
* Add custom time fields feature to the web_log collector.
* Add a Go version of the PowerDNS Authoritative Nameserver
collector (powerdns).
* Add a Go version of the PowerDNS Recursor collector
(powerdns_recursor).
* Add a Go version of the PowerDNS DNSdist collector (dnsdist).
* Add a Dnsmasq DNS Forwarder collector (dnsmasq).
* Add collecting directories size to the filecheck collector.
* Add old systemd versions support to the systemdunits collector.
* Add unmatched lines logging to the web_log collector.
Notifications
* Add API V2 support to the PagerDuty health integration.
* Add threads support to the Google Hangouts health integration.
* Add a Opsgenie health integration.
Exporting
* Add HTTP and HTTPS support to the simple exporting connector.
Packaging/installation
* Update React dashboard v2.11.
* Update go.d.plugin version to v0.26.2.
* Update eBPF collector to 0.4.9.
* Add ability to use system libwebsockets instead of bundled
version.
* Update the version of libJudy that we bundle to 1.0.5-netdata2.
Bug fixes
* Fix crash in the eBPF plugin by initializing variables.
* Fix sending chart definition on every data collection in alarms
collector.
* Fix a lock check.
* Fix issue with chart metadata sent multiple times over ACLK.
* Fix a buffer overflow when extracting information from a
streaming connection.
* Fix hostname configuration in the exporting engine.
* Fix units and data source exporting options.
* Fix exporting config.
* Fix health by disabling used_file_descriptors alarm.
* Fix GPU data filtering in the nvidia_smi collector.
* Fix username resolution in the nvidia_smi collector.
* Fix compilation with HTTPS disabled.
* Fix hostname when syslog is used in syslog health integration.
* Fix streaming buffer size.
* Fix database endless loop when cleaning obsolete charts.
* Disable chart obsoletion code for archived chart creation.
* Fix Prometheus remote write exporter so that it doesn't stop
when data is not available for dimension formatting.
* Fix memory calculation by moving shared from cached to used
dimension.
* Fix parsing in the libreswan collector.
* Fix an infinite loop in the statsd plugin
* Disregard host tags configuration pointer.
* Fix platform dependent printf format.
* Fix cgroups collector resolving container names in k8s.
* Fix python.d plugin runtime chart creation.
* Fix race condition in rrdset_first_entry_t() and
rrdset_last_entry_t().
* Fix the data endpoint so that the context param is correctly
applied to children.
* Fix Coverity errors (CID 364045,364046).
* Fix the elasticsearch_last_collected alarm.
* Fix spelling error in xenstat.plugin.
* Fix chart filtering.
* Fix libnetdata headers to be compatible with C++.
* Fix registry responses to remove caching.
* Fix eBPF memory management.
* Fix overlapping memory issue.
* Fix response and upstream response time histogram charts in the
web_log collector.
* Fix logs timestamps always in UTC issue in the go.d.plugin
* Fix collecting slave status for MariaDB v10.2.0- in the mysql
collector
* Fix cumulative_stats configuration option in the unbound
collector
* Fix parsing configuration file (respect 'include-toplevel'
directive) in unbound collector
* Fix handling charts with type.id >= 200 (netdata limit) in
go.d.plugin
* Fix parsing version query response in the mysql collector
* Fix Netsplits chart dimensions algorithm in the the vernemq
collector.
* Fix a typo in dashboard_info.js for VerneMQ.