Real-time performance monitoring
netdata is a highly optimized Linux daemon providing real-time performance monitoring for Linux systems, Applications, SNMP devices, over the web!
- Developed at network
- Sources inherited from project openSUSE:Factory
-
5
derived packages
- Download package
-
Checkout Package
osc -A https://api.opensuse.org checkout openSUSE:42:Factory-Candidates-Check/netdata && cd $_
- Create Badge
Refresh
Refresh
Source Files
Filename | Size | Changed |
---|---|---|
netdata-logrotate-su.patch | 0000000252 252 Bytes | |
netdata-rpmlintrc | 0000000095 95 Bytes | |
netdata-smartd-log-path.patch | 0000000660 660 Bytes | |
netdata-v1.15.0.tar.gz | 0003662255 3.49 MB | |
netdata.changes | 0000050208 49 KB | |
netdata.spec | 0000005108 4.99 KB |
Revision 10 (latest revision is 58)
Dominique Leuenberger (dimstar_suse)
accepted
request 705267
from
Mia Herkt (lachs0r)
(revision 10)
- Update to v1.15.0 Bug Fixes: * Prowl notifications were not being sent, unless another notification method was also active * Fix exception handling in the python.d plugin * The node applications group did not include all node processes. * The nvidia_smi collector displayed incorrect power usage * The python.d plugin would sometimes hang, because it lacked a connect timeout * The mongodb collector raised errors due to various KeyErrors * The smartd_log collector would show incorrect temperature values Improvements: * Support for aggregate node view * Database engine * New collector modules: + Go.d collectors for OpenVPN, the Tengine web server and ScaleIO (VxFlex OS) instances + Monitor disk access latency like ioping does * Energi Core daemon monitoring, suits other Bitcoin forks * Collector improvements: + Add docker swarm manager metrics to the go.d docker_engine collector + Implement unified cgroup cpu limit + python.d.plugin: Allow monitoring of HTTP(S) endpoints which require POST data and make the UrlService more flexible * Support the AWS Kinesis backend for long-term storage * Add a new "text-only" chart renderer * API Improvements: + Smarter caching of API calls. Do not cache alarms and info api calls and extend no-cache headers. + Extend the api/v1/info call response with system and collector information + k6 script for API load testing * Kubernetes helmchart improvements: + Added the init container, where sysctl params could be managed, to bypass the Cannot allocate memory issue + Better startup/shutdown of slaves and reduced memory usage with liveness/readiness probes and default memory mode none + Added the option of overriding the default settings for kubelet, kubeproxy and coredns collectors via values.yaml + Make the use of persistent volumes optional, add apiVersion to fix linting errors and correct the location of the env field - Update to v1.14.0 The release introduces major additions to Kubernetes monitoring, with tens of new charts for Kubelet, kube-proxy and coredns metrics, as well as significant improvements to the netdata helm chart. Two new collectors were added, to monitor Docker hub and Docker engine metrics. Finally, v1.14 adds support for version 2 cgroups, OpenLDAP over TLS, NVIDIA SMI free and per process memory and configurable syslog facilities. Bug Fixes: * Fixed problem autodetecting failed jobs in python.d plugin. It now properly restarts jobs that are being rechecked, as soon as they are able to run. * CouchdDB monitoring would stop sometimes with an exception. Fixed the unhandled exception causing the issue. * The netdata api deliberately returned http error 400 when netdata ran in memory mode none. Modified the behavior to return responses, regardless of the memory mode * The python.d plugin sometimes does not receive SIGTERM when netdata exits, resulting in zombie processes. Added a heartbeat so that the process can exit on SIGPIPE. * The new SMS Server Tools notifications did not handle errors well, resulting in cryptic error messages. Improved error handling. * Fix segmentation fault in streaming, when two dimensions had similar names. * Kubernetes Helm Chart: Fixed incorrect use of namespaces in ServiceAccount and ClusterRoleBinding RBAC fixes. * Elastic search: The option to enable HTTPS was not included in the config file, giving the erroneous impression that HTTPS was not supported. The option was added. * RocketChat notifications were not being sent properly. Added default recipients for roles in the health alarm notification configuration. Improvements: * go.d.plugin v0.4.0 : Docker Hub and k8s coredns collectors, springboot2 URI filters support. * go.d.plugin v0.3.1 : Add default job to run k8s_kubelet.conf, k8s_kubeproxy, activemq modules * go.d.plugin v0.3.0 : Docker engine, kubelet and kub-proxy collectors. x509check module reading certs from file support * Added unified cgroup support that includes v2 cgroups * Disk stats: Added preferred disk id pattern, so that users can see the id they prefer, when multiple ids appear for the same device * NVIDIA SMI: Added memory free and per process memory usage charts to the collector * OpenLDAP: Added TLS support, to allow monitoring of LDAPS. * PHP-FPM: Add health check to raise alarms when the phpfm server is unreachable * PostgreSQL: Our configuration options to connect to a DB did not support all possible option. Added option to connect to a PostreSQL instance by defining a connection string (URI). * python.d.plugin: There was no way to delete obsolete dimensions in charts created by the python.d plugin. The plugin can now delete dimension at runtime. * netdata supports sending its logs to Syslog, but the facility was hard-coded. We now support configurable Syslog facilities in netdata.conf. * Kubernetes Helm Chart improvements: + Added serviceName in statefulset spec to align with the k8s documentation + Added preStart command to persist slave machine GUIDs, so that pod deletion/addition during upgrades doesn't lose the slave history. + Disabled non-essential master netdata collector plugins to avoid duplicate data + Added preStop command to wait for netdata to exit gracefully before removing the container + Extended configuration file support to provide more control from the helm command line + Added option to disable Role-based access control + Added liveness and readiness probes. - Update to v1.13.0 netdata has taken the first step into the world of Kubernetes, with a beta version of a Helm chart for deployment to a k8s cluster and proper naming of the cgroup containers. We have big plans for Kubernetes, so stay tuned! A major refactoring of the python.d plugin has resulted in a dramatic decrease of the required memory, making netdata even more resource efficient. We also added charts for IPC shared memory segments and total memory used. Improvements: * Kubernetes: Helm chart and proper cgroup naming * python.d.plugin: Reduce memory usage with separate process for initial module checking and loaders cleanup * IPC shared memory charts * mysql module add ssl connection support * FreeIPMI: Have the debug option apply the internal freeipmi debug flags * Prometheus backend: Support legacy metric names for source=avg * Registry: Allow deleting the host we are looking at * SpigotMC: Use regexes for parsing. Bug Fixes: * Postgres: fix connection issues * Proxmox container: Fix cgroup naming and use total_* memory counters for cgroups * proc.plugin and plugins.d: Fix memory leaks * SpigotMC: Fix UnicodeDecodeError and py2 compatibility fix * Fix non-obsolete dimension deletion * UI: Fix incorrect icon for the streaming master * Docker container names: Retry renaming when a name is not found * apps.plugin: Don't send zeroes for empty process groups * go.d.plugin: Correct sha256sum check * Unbound module: Documentation corrected with troubleshooting section. * Streaming: Prevent UI issues upon GUID duplication between master and slave netdata instances * Linux power supply module: Fix missing zero dimensions * Minor fixes around plugin_directories initialization - Update to v1.12.2 Bug Fixes: * Zombie processes exist after restart netdata - add heartbeat to python.d plugin * RocketChat notifications not working * SIGSEGV crash during shutdown of tc plugin * CMake warning for nfacct plugin Improvements: * Oracledb python module * Show streamed servers even for users that are not signed in - Drop GPG signature (no longer used) - Drop spec compatibility with old distro versions - Drop netdata-automake-no-dist-xz.patch - Refresh netdata-smartd-log-path.patch - Update to v1.12.1 Fixes: * Fix SIGSEGV at startup: Don't free vars of charts that do not exist #5455 * Prevent invalid Linux power supply alarms during startup #5447 * Correct duplicate flag enum in health.h #5441 * Remove extra 'v' for netdata version from Server response header #5440 and spec URL #5427 * apcupsd - Treat ONBATT status the same as ONLINE #5435 * Fix #5430 - LogService._get_raw_data under python3 fails on undecodable data #5431 * Correct version check in UI #5429 * Fix ERROR 405: Cannot download charts index from server - cpuidle handle newlines in names #5425 * Fix clock_gettime() failures with the CLOCK_BOOTTIME argument #5415 * Use netnsid for detecting cgroup networks; #5413 * Python module sensors fix #5406 * Fix ceph.chart.py for Python3 #5396 (GaetanF) * Fix warning condition for mem.available #5353 * cups.plugin: Support older versions #5350 Improvements: * Add driver-type option to the freeipmi plugin #5384 * Add support of tera-byte size for Linux bcache. #5373 * Split nfacct plugin into separate process #5361 * Add cgroup cpu and memory limits and alarms #5172 * Add message queue statistics #5115 - Update to v1.12.0 Key improvements: * Introducing netdata.cloud, the free netdata service for all netdata users * High performance plugins with go.d.plugin (data collection orchestrator written in Go) * 7 new data collectors and 11 rewrites of existing data collectors for improved performance * A new management API for all netdata servers * Bind different functions of the netdata APIs to different ports Management API: Netdata now has a management API. We plan to provide a full set of configuration commands using this API. In this release, the management API supports disabling or silencing alarms during maintenance periods. For more information about the management API, check https://docs.netdata.cloud/web/api/health/#health-management-api Anonymous statistics: Anonymous usage information can be collected and sent to Google Analytics. This functionality is disabled by default in openSUSE. Remove /etc/netdata/.opt-out-from-anonymous-statistics to enable. The statistics calculated from this information will be used for: 1. Quality assurance, to help us understand if netdata behaves as expected and help us identify repeating issues for certain distributions or environment. 2. Usage statistics, to help us focus on the parts of netdata that are used the most, or help us identify the extend our development decisions influence the community. Information is sent to Netdata via two different channels: * Google Tag Manager is used when an agent's dashboard is accessed. * The script anonymous-statistics.sh is executed by the Netdata daemon, when Netdata starts, stops cleanly, or fails. Both methods are controlled via the same opt-out mechanism. For more information, check https://docs.netdata.cloud/docs/anonymous-statistics/ Data collection: This release introduces a new Go plugin orchestrator. This plugin has its own github repo (https://github.com/netdata/go-orchestrator). It is open-source, using the same license and we welcome contributions. The orchestrator can also be used to build custom data collection plugins written in Go. We have used the orchestrator to write many new Go plugins in our go.d plugin github repo. For more information, check https://github.com/netdata/go-orchestrator#go-orchestrator-wip New data collectors: * Activemq (Go) * Consul (Go) * Lighttpd2 (Go) * Solr (Go) * Springboot2 (Go) * mdstat - nonredundant arrays (C) * CUPS printing system (C) High performance versions of older data collectors: * apache (Go) * dns_query (Go) * Freeradius (Go) * Httpcheck (Go) * Lighttpd (Go) * Portcheck (Go) * Nginx (Go) * cpufreq (C) * cpuidle (C) * mdstat (C) * power supply (C) Other improved data collectors: * Fix the python plugin clock (collectors falling behind). * adaptec_raid: add to python.d.conf. * apcupsd: Detect if UPS is online. * apps: Fix process statistics collection for FreeBSD. * apps: Properly lookup docker container name when running in ECS * fail2ban: Add 'Restore Ban' action. * go_expavar: Don't check for duplicate expvars. * hddtemp: Don't use disk model as dim name. * megacli: add to python.d.conf. * nvidia_smi: handle N/A values. * postgres: Fix integer out of range error on Postgres 11, fix locks count. * proc: Don't show zero charts for ZFS filesystem. * proc; Fix cached memory calculation. * sensors: Don't ignore 0 RPM fans on start. * smartd_log: check() unhandled exception: list index out of range. * SNMP: Gracefully ignore the offset if the value is not a number Health Monitoring: * Add Prowl notifications for iOS users. * Show count of active alarms per state in email notifications. * Show evaluated expression and expression variable values in email notifications. * Improve support for slack recipients (channels/users). * Custom notifications: Fix bug with alarm role recipients. Dashboards: * Server filtering in my-netdata menu when signed in to netdata.cloud * All units are now IEC-compliant abbreviations (KiB, MiB etc.). * GUI: Make entire row clickable in the registry menu showing the list of servers. Backends: * Do not report stale metrics to prometheus. Other: * Treat DT_UNKNOWN files as regular files. * API: Stricter rules for URL separators. - Update to v1.11.1 Improved internal database: Overflown incremental values (counters) do not show a zero point at the charts. Netdata detects the width (8bit, 16bit, 32bit, 64bit) of each counter and properly calculates the delta when the counter overflows. The internal database format has been extended to support values above 64bit. New data collection plugins: * openldap, to collect performance statistics from OpenLDAP servers. * tor, to collect traffic statistics from Tor. * nvidia_smi to monitor NVIDIA GPUs. Improved data collection plugins: * BUG FIX: network interface names with colon (:) in them were incorrectly parsed and resulted in faulty data collection values. * BUG FIX: smartd_log has been refactored, has better python v2 compatibility, and now supports SCSI smart attributes * cpufreq has been re-written in C - since this module if common, we decided to convert to an internal plugin to lower the pressure on the python ones. There are a few more that will be transitioned to C in the next release. * BUG FIX: sensors got some compatibility fixes and improved handling for lm-sensors errors. Health monitoring: * BUG FIX: max network interface speed data collection was faulty, which resulted in false-positive alarms on systems with multiple interfaces using different speeds (the speed of the first network interface was used for all network interfaces). Now the interface speed is shown as a badge. * alerta.io notifications got a few improvements * BUG FIX: conntrack_max alarm has been restored (was not working due to an invalid variable name referenced) Registry (my-netdata menu): * It has been refactored a bit to reveal the URLs known for each node and now it supports deleting individual URLs. - Update to 1.11.0 * Stock config files are now in /usr/lib/netdata; use the /etc/netdata/edit-config script to copy and edit them. * The query engine of netdata has been re-written to support query plugins. We have already added the following algorithms that are available for alarm, charts and badges: + stddev, for calculating the standard deviation on any time-frame. + ses or ema or ewma, for calculating the exponential weighted moving average, or single/simple exponential smoothing on any time-frame. + des, for calculating the double exponential smoothing on any time-frame. + cv or rsd, for calculating the coefficient of variation for any time-frame. Fixed security issues: * CVE-2018-18836 Fixed JSON Header Injection (an attacker could send \n encoded in the request to inject a JSON fragment into the response). * CVE-2018-18837 Fixed HTTP Header Injection (an attacker could send \n encoded in the request to inject an HTTP header into the response). * CVE-2018-18838 Fixed LOG Injection (an attacker could send \n encoded in the request to inject a log line at access.log). * CVE-2018-18839 Not fixed Full Path Disclosure, since these are intended (netdata reports the absolute filename of web files, alarm config files and alarm handlers). * Fixed Privilege Escalation by manipulating apps.plugin or cgroup-network error handling. * Fixed LOG injection (by sending URLs with \n in them). New data collection modules: * rethinkdbs for monitoring RethinkDB performance * proxysql for monitoring ProxySQL performance * litespeed for monitoring LiteSpeed web server performance. * uwsgi for monitoring uWSGI performance * unbound for monitoring the performance of Unbound DNS servers. * powerdns for monitoring the performance of PowerDNS servers. * dockerd for monitoring the health of dockerd * puppet for monitoring Puppet Server and Puppet DB. * logind for monitoring the number of active users. * adaptec_raid and megacli for monitoring the relevant raid controller * spigotmc for monitoring minecraft server statistics * boinc for monitoring Berkeley Open Infrastructure Network Computing clients. * w1sensor for monitoring multiple 1-Wire temperature sensors. * monit for collecting process, host, filesystem, etc checks from monit. * linux_power_supplies for monitoring Linux Power Supplies attributes Data collection orchestrators changes: * node.d.plugin does not use the js command any more. * python.d.plugin now uses monotonic clocks. There was a discrepancy in clocks used in netdata that resulted in a shift in time of python module after some time (it was missing 1 sec per day). * added MySQLService for quickly adding plugins using mysql queries. * URLService now supports self-signed certificates and supports custom client certificates. * all python.d.plugin modules that require sudo to collect metrics, are now disabled by default, to avoid security alarms on installations that do not need them. Improved data collection modules: * apps.plugin now detects changes in process file descriptors, also fixed a couple of memory leaks. Its default configuration has been enriched significantly, especially for IoT. * freeipmi.plugin now supports option ignore-status to ignore the status reported by given sensors. * statsd.plugin (for collecting custom APM metrics) + The charting thread has been optimized for lowering its CPU consumption when several millions of metrics are collected. + sets now report zeros instead of gaps when no data are collected + histograms and timers have been optimized for lowering their CPU consumption to support several thousands of such metrics are collected. + histograms had wrong sampling rate calculations. + gauges now ignore sampling rate when no sign is included in the value. + the minimum sampling rate supported is now 0.001. + netdata statsd is now drop-in replacement for datadog statsd (although statsd tags are currently ignored by netdata). * proc.plugin (Linux, system monitoring) + Unused interrupts and softirqs are not used in charts (this saves quite some processing power and memory on systems with dozens of CPU cores). + fixed /proc/net/snmp parsing of IcmpMsg lines that failed on a few systems. + Veritas Volume Manager disks are now recognized and named accordingly. + Now netdata collects TcpExtTCPReqQFullDrop and re-organizes metrics in charts to properly monitor the TCP SYN queue and the TCP Accept queue of the kernel. + Many charts that were previously reported as IPv4, were actually reflecting metrics for both IPv4 and IPv6. They have been renamed to ip.*. + netdata now monitors SCTP. + Fixed BTRFS over BCACHE sector size detection. + BCACHE data collection is now faster. + /proc/interrupts and /proc/softirqs parsing fixes. * diskspace.plugin (Linux, disk space usage monitoring) + It does not stat() excluded mount points any more (it was interfering with kerberos authenticated mount points). + several filesystems are now by default excluded from disk-space monitoring, to avoid breaking suspend on workstations. * python.d.plugin PYTHON modules (applications monitoring) + web_log module now supports virtual hosts, reports http/https metrics, support squid logs + nginx_plus module now handles non-continuous peer IDs + ipfs module is optimized, the use of its Pin API is now disabled by default and can enabled with a netdata module option (using the IPFS Pin API increases the load on the IPFS server). + fail2ban module now supports IPv6 too. + ceph module now checks permissions and properly reports issues + elasticsearch module got better error handling + nginx_plus module now uses upstream ip:port instead of transient id to identify dimensions. + redis, now it supports Pika, collects evited keys, fixes authentication issues reported and improves exception handling. + beanstalk, bug fix for yaml config loading. + mysql, the % of active connections is now monitored, query types are also charted. + varnish, now it supports versions above 5.0.0 + couchdb + phpfpm, now supports IPv6 too. + apache, now supports IPv6 too. + icecast + mongodb, added support for connect URIs + postgress + elasticsearch, now it supports versions above 6.3.0, fixed JSON parse errors + mdstat , now collects mismatch_cnt + openvpn_log * node.d.plugin NODE.JS modules + snmp was incorrectly parsing a new OID names as float. * charts.d.plugin BASH modules + nut now supports naming UPSes. Health monitoring: * Added variable $system.cpu.processors. * Added alarms for detecting abnormally high load average. * TCP SYN and TCP accept queue alarms, replacing the old softnet dropped alarm that was too generic and reported many false positives. * system alarms are now enabled on FreeBSD. * netdata now reads NIC speed and sets alarms on each interface to detect congestion. * Network alarms are now relaxed to avoid false positives. * New bcache alarms. * New mdstat alarms. * New apcupsd alarms. * New mysql alarms. * New notification methods: + rocket.chat + Microsoft Teams + syslog + fleep.io + Amazon SNS Backends: * Host tags are now sent to Graphite * Host variables are now sent to Prometheus Streaming: * Each netdata slave and proxy now filter the charts that are streamed. This allows exposing netdata masters to third parties by limiting the number of charts available at the master. * Fixed a bug in streaming slaves that randomly prevented them to resume streaming after network errors. * Fixed a bug that on slaves that sent duplicated chart names under certain conditions. * Fixed a bug that caused slaves to consume 100% CPU (due to a misplaced lock) when multiple threads were adding dimensions on the same chart. * The receiving nodes of streaming (netdata masters and proxies) can now rate-limit the rate of inbound streaming requests received. * Re-worked time synchronization between netdata slaves and masters. API: * Badges that report time, now show "undefined" instead of "never". Dashboard: * Added UTC timezone to the list of available time-zones. * The dashboard was sending some non-HTTP compliant characters at the URLs that made netdata dashboards break when used under certain proxies. - Remove _service - Add GPG signature - Use %license macro - Update to v1.10.0 * new plugins: - BTRFS - monitor the allocations of BTRFS filesystems (netdata can now properly detect when btrfs is going out of space) - BCACHE - monitor the caching block layer that allows building hybrid disks using normal HDDs and SSDs - Ceph - monitor ceph distributed storage - nginx plus - monitor the nginx+ web servers - libreswan - monitor IPSEC tunnels - Traefik - monitor traefik reverse proxies - icecast - monitor icecast streaming servers - ntpd - monitor NTP servers - httpcheck - monitor any remote web server - portcheck - monitor any remote TCP port - spring-boot - monitor java spring boot applications - dnsdist - monitor dnsdist name servers - hugepages - monitor the allocation of Linux hugepages * enhanced/improved plugins: - statsd + statsd dimensions now support the options the external plugin dimensions support (currently the only usable option is hidden to add the dimension, but make it hidden on the dashboard - a hidden dimension can participate in various calculations, including alarms). + statsd now reports the CPU usage of its threads at the netdata section. + statsd metrics are logged to access.log the first time they are encountered. + statsd metrics now accept the special value zinit to allow them get initialized without altering their values (this is useful if you have rare metrics that you need to initialize when netdata starts). + statsd over TCP is now a lot faster - netdata can process up to 3.5mil statsd metrics / second using just one core. Added options to control the timeouts of TCP statsd connections. + fixed the title and context of statsd private charts + statsd private charts can now be hidden from the dashboard - web_log + Added web server response timings histogram - containers monitoring + netdata now monitors systemd-nspawn containers. + netdata now renames charts of kubernetes containers. + virsh is now called with -r to avoid prompting for password + cgroup-network is now a lot more strict, preventing unauthorized privilege escalation + cgroup-network now searches for container processes in sub-cgroups too - this improves the mapping of network interfaces to containers + cgroup-network now works even when there are no veth interfaces in the system - system memory + treat slab memory as cached + added a new chart for monitoring the memory available for use, before hitting swap + netdata now monitors Linux hugepages and (transparent) hugepages - diskspace + support huge amounts of mountpoints - netdata was crashing with stack overflow due to recursion - now it is a loop, so any number of mount points is supported - network interfaces + moved tcp passive and active opens to a separate chart, to allow the TCP issues dimensions scale better by default + updated the information presented on TCP charts to match the latest v4.15 kernel source - postgres + new charts: checkpointer bgwriter autovacuum replication delta WAL archive WAL temporary files + The postgres plugin now also works when postgres is in recovery mode. - rabbitmq + added Erlang run queue chart. This is useful in conjunction with the existing Erlang processes chart to get a better overall idea of what's going on in the Erlang VM. + added rabbitmq information on the dashboard to complement the charts. - apps.plugin netdata prior to this version was detecting the user and group of processes by examining the ownership of /proc/PID/stat. Unfortunately it seems that the owneship of files in /proc do not change when the process switches user. So, netdata could not detect the user and group of processes that started as root and then switched to another user. Now netdata reads /proc/PID/status: + process ownship information is now accurate + eliminated the need to read /proc/PID/statm (all the information of /proc/PID/statm is available in /proc/PID/status) + allowed netdata to read VmSwap, so a new chart has been added to monitor the swap memory usage per process, user and group. + The new plugin is 20% more expensive in terms of CPU. We tried hard to optimize it, but this is as good as it can get. - haproxy + hrsp_1xx, hrsp_2xx, hrsp_3xx, hrsp_4xx, hrsp_5xx, hrsp_other, hrsp_total for backands and frontends + qtime, ctime, rtime, ttime metrics for backend servers + backend servers In UP state - uptime + netdata now uses /proc/uptime when CLOCK_BOOTTIME does not report the same uptime. In containers CLOCK_BOOTTIME reports the uptime of the host, while /proc/uptime reports the uptime of the container, so now netdata correctly reports the uptime of the container. - ksm + removed to_scan dimension + the savings % reported by netdata was less than the actual - fixed it. - mdstat + various fixes to better monitor rebuild time and rate - elasticsearch + Added several charts for translog / indices segments statistics and JVM buffer pool utilization, which are often helpful when evaluating an elasticsearch node health - apcupsd + netdata now supports monitoring multiple APC UPSes. - isc-dhcpd + netdata now also supports monitoring IPv6 leases - fronius + added a new dimension solar_consumption + added alarms - stiebeleltron + added alarms - python.d + python.d.plugin can now start even if /etc/netdata/python.d.conf is missing + python.d.plugin now has an internal run counter + the unicode decoding of the plugin has been fixed + the plugin now does not validate self-signed certificates + the plugin can not revive obsolete charts - charts.d + charts.d.plugin BASH modules can now have custom number of retries in case of data collection failures * new alarm notifications plugins: - alerta - IRC * web server: - netdata now has a new internal web server that supports a fixed number of threads - we call it static web server. This web server allows netdata to work around memory fragmentation (since the treads are fixed, the underlying memory allocators reuse the same memory arenas) and cpu utilization (we can control the number of threads that will be used by netdata). This is the default now. - now the static threads web server reports the CPU usage of each of its threads. - the HTTP response headers now include the netdata version * dashboard: - the print button now respects the URL path netdata is hosted. - dygraphs updated to the latest version - this fixes an issue that prevented netdata charts from being interactive under certain conditions - added dygraph theme logscale - fontawesome updated to version 5 - d3 updated to the latest version (this broke c3 charts that require an older version) - added d3pie charts - custom dashboards can now have alarms for specific roles (all, none, one or more). - allow stacked charts to zoom vertically when dimensions are selected - netdata now has a global XSS protection - netdata now uses intersectionObserver when available - this improves the scrolling performance of the dashboard. - prevent date, time and units from wrapping at the charts legends - various units scaling improvements - added data-common-colors="NAME" chart option for custom dashboards - added wiki page for creating custom dashboards on Atlassian's Confluence - prevented a double click on the charts' toolbox to select the text of the buttons. - fixed the alignment of dashboard icons - added a simple js, called refresh-badges.js, to update badges on a custom web page * badges: - netdata badges can now be scaled * API: - added gtime parameter, for group time. This is used to request from netdata to return values in a different rate (i.e. gtime=60 on a X/sec dimension, will return X/min). - fixed a rounding bug in JSON generation - the dimensions= parameter now supports simple patterns and added option values match-ids and match-names to control which matches are executed for dimensions. * alarms: - "system.swap" alarms now send notifications with a 30 seconds delay, to work-around a kernel bug that incorrectly reports all swap as instantly used under containers - added alarm to predict the time a mount point will run out of inodes - added alerta.io notifications - added available memory alarm - removed unsupported html tags from hipchat notifications. - pagerduty notifications have been modified to avoid incident duplication - alarm definitions can now use both chart IDs and chart names (prior to this version only chart IDs were allowed). - curl options (eg for disabling SSL certificates verification) for alarm-notify.sh can now be defined in health_alarm_notify.conf. - netdata can now send notifications to IRC channels * backends: - on netdata masters, allow filtering the hosts that will be sent to backends with send hosts matching = * pattern. - improved connection error handling and added retries to allow netdata connect to certain backends that failed with EALREADY or EINPROGRESS. - json backends now receive host tags (the tags have to be formatted in a json friendly way) - re-worked the alarm that triggers when backend data are lost, to avoid flip-flops. - prometheus backends: + added URL option timestamps=yes|no to /api/v1/allmetrics to support prometheus Pushgateway + added netdata_info variable with the version of netdata + renamed netdata_host_tags to netdata_host_tags_info (the old exists but is deprecated and will be removed eventually) + when prometheus uses average metrics, netdata remembers the last access time the prometheus collected metrics, on a per host basis. * metrics streaming between netdata: - netdata masters and proxies now expose the version of the netdata collecting the metrics, not their own. So, now a netdata master shows on the dashboard and sends to backends the version of the netdata collecting the metrics - added stream.conf option multiple "connections = accept | deny" to allow or deny multiple connection for the same netdata host. The default remains "accept", but it is likely to be changed to no on future versions. * other: - added global option gap when lost iterations to control the number of iterations that should be lost to show a gap on the charts. - various fixes/improvements related to netdata logs - the main change is that now netdata logs the thread name that logged the message, providing helpful insights about the thread that complained. - re-worked the exit procedure of netdata to allow it cleanup properly - sometimes netdata was deadlocked during exit, waiting forever - now netdata always exits promptly - fixed compilation on ancient gcc versions - netdata was always setting itself to the idle process scheduling priority, even when it was configured to do otherwise. Fixed it. - Update to v1.9.0 * new features: - IP address ACLs - auto-scaling units (KB → MB → GB → TB etc.) - dashboard snapshots for any timeframe/resolution - detection of TCP overflows and dropped connections - detection of redis background save failures - export dashboards to PDF - flock.com and kavenegar.com alarm notifications - highlighting timeframes on all charts by holding Ctrl or Alt - monitoring of couchdb, powerdns, beanstalkd and dnsdist - monitoring of libvirt VMs - new alarms: 1. too many tcp orphan sockets 2. tcp memory that detects that the tcp stack is under memory pressure or close to giving memory errors 3. too many tcp connections (for kernels that do not support dynamic allocation of connections) - timezone support - unit conversion (temperatures, time, etc.) * fixes/improvements: - better/faster dashboard scrolling - python.d.plugin rewritten + supports option autodetection_retry: SECONDS + the new URL service disables certificates checks by default, to allow self-signed certificates to work without configuration. - streaming: + netdata proxies with more than 100 slaves had a timing issue that caused them to crash randomly on slave reconnects. Parts of the code have been rewritten to get rid of the timing issue. + netdata slaves and proxies now have a protection that ensures they will never use 100% CPU, even if the master is misbehaving. + expired orphaned hosts are now removed from the my-netdata menu of the dashboard. + streaming functions can now be monitored via access.log + streaming now transfers alarm variables too - web_log plugin + custom regex now supports parsing hostnames and IPs + now parses lines with error 408 (request timeout - these are a special case, since the request has not received by the web server, so the log line is incomplete) + now properly parses resp_length with value - disks under Linux are renamed using /dev/disk/by-label. An option has been added at netdata.conf to also allow renaming based on /dev/disk/by-id. - chrony is now disabled by default, because there have been reports that chronyc enters an infinite loop in CentOS and RHEL. - tomcat improvements to support flavors of the tomcat server - chart system.io (the total system Disk I/O) is now calculated by aggregating the reads and writes of all physical disks. The previous system.io chart (that is based on pgpgin and pgpgout from /proc/vmstat) is now named system.pgpgio. The key difference is that the new system.io now sees ZFS I/O, and it also correctly and accurately sums the real disk bandwidth of RAID arrays. - chart system.net (the total system network bandwidth) is now calculated by aggregating the bandwidth of all physical network interfaces and is common for both IPv4 and IPv6. - tc (QoS) charts now sort the dimensions on the legends, the same way tc reports them. - mysql (and mariadb) got new charts for galera replication - postgres versions <= 10 the WAL directory was named pg_xlog' and from 10 upwards has been renamed to pg_wal - freeipmi.plugin got a command line option (can be given in netdata.conf) to ignore certain sensor IDs that are faulty. - minor openvpn_log, smartd, smartd_log, varnish, mdstat, dns_query_time, isc_dhcpd, freeradius and node.d.plugin fixes/improvements - netdata now supports multiple plugin directories. The setting is the same in netdata.conf, plugins directory = "DIRECTORY1" "DIRECTORY2" ..., up to 20 directories. - netdata now supports alarms variables. Each plugin can now define host global and chart local variables with static values that can be used in alarms' expressions. - dashboard: + hover selection of charts is now faster on all browsers + the dashboard is now fixed when a modal is open, preventing scrolling + the dashboard now uses fontawesome 5.0.1 for icons + chart names can now be searched with browser control-F + netdata now detects libvirt VM network interfaces and moves them to the VM section of the dashboard + now shows the context, resolution and plugin/module of each chart in tooltips + should now put a lot less CPU pressure on the browser when the page does not have focus - statsd: + metrics can now be added to statsd synthetic charts using patterns + dimensions added to statsd synthetic charts can automatically be renamed using a dictionary + timers and histograms now report zeros when nothing is collected - badges: + fixed a bug in netdata badges that was incorrectly matching zero values with the null color condition + added API option display_absolute to allow badges use the signed value for color evaluation, but present the absolute value - alarms/notifications: + warning emails sent by netdata are now a little bit more orange (they were a bit greenish) + fixed a bug in email notifications that was triggering a corrupted MIME match by anti-spam solutions + pushbullet notifications now track the devices, so that per-device filtering at pushbullet is possible. Also improved the formatting a bit. + correct priority of warnings for pushover notifications + alarms can now use variables like this: ${variable with spaces or +, -, *, / in it} - other: + access.log has been refactored to support monitoring all netdata operations + inodes monitoring is now by default disabled for mount points based on filesystems that do not have a maximum inode threshold (such as cephfs) + rabbitmq has been added to apps_groups.conf so that apps.plugin now monitors (cpu, memory, disk I/O, sockets, etc) rabbitmq instances + several email and log management apps have been added to email and logs targets of apps_groups.conf + ceph target added to apps_groups.conf to allow netdata to monitor Ceph + refactored several internal data collection plugins to eliminate a few hundreds of index lookups per second + netdata.conf settings that are loaded from disk, but were the same with the default ones, were generated commented when the server was asked to return its config. Now all loaded settings are generated uncommented. + netdata simple patterns can now extract the the wildcarded part of the string they match (used in statsd synthetic charts) + netdata simple patterns can allow escaping spaces by prefixing them with a backslash - Update netdata-smartd-log-path.patch - Remove netdata-main-return.patch (fixed upstream) - Fix rpmlint env-script-interpreter warnings - Add return statement to main function (-Wreturn-type) * netdata-main-return.patch - Update to v1.8.0 * fixed bugs: - streaming slaves consuming 100% CPU - missing alarm notifications on netdata masters - API generating corrupted JSON - crashing when starting on systems without writable disks - python.d.plugin URLService did not support HTTP keep-alive - streamed charts with duplicate names * netdata enhancements: - netdata can now listen on UNIX domain sockets - systemd netdata.service now allows setting negative netdata OOM score and restarts netdata if it crashes * new plugins: - CPU thermal throttling charts - chrony plugin - Stiebel Eltron plugin to collect metrics from heat pumps and hot water installations from Stiebel Eltron ISG * improved plugins: - web_log bugfixes, enhancements and optimizations (including squid logs) - web_log now enables parsing HTTP/2 logs in custom_log_format - redis bugfixes - haproxy bugfixes - elasticsearch bugfixes and optimizations - rabbitmq bugfixes and optimizations - mdstat bugfixes - tomcat improvements - mysql improvements - dovecot improvements - postgres improvements - cpufreq fixed a bug that prevented accurate reporting of CPU frequencies. - cpuidle performance improvements (faster under load) - fail2ban bugfixes - SNMP plugin new uses latest net-snmp and the corrupted 64-bit counters encountered under certain node.js versions is fixed * dashboard improvements: - easypiecharts and gauges can now render arbitrary ranges and animate clockwise or counterclockwise - container network interfaces are now moved to the container section - containers and VMs now have summary gauges on the dashboard - traditionally netdata was using 1024 bits = 1 kilobit. It is fixed: 1000 bits = 1 kilobit. - netdata charts should now work on wordpress pages * alarms and notifications: - alarm-notify.sh now supports debug mode, showing the exact commands it runs to send notifications, when export NETDATA_ALARM_NOTIFY_DEBUG=1 - alarm-notify.sh now supports setting the sender email address of the emails it sends - emails sent by alarm-notify.sh now include headers to reduce the possibility of them being scored as spam - network related alarms got new thresholds and improved badges - netdata now detects if the system has been suspended and pauses all alarms for 60 seconds on resume, to prevent false alarms - netdata alarms now support filtering based on hostname and OS - slack notifications now show the host that sent the alarm * statsd: - the number of fractional points supported by statsd is now configurable (1 to 7) - fixed: 95th percentile calculation on statsd histograms and timers was incorrectly averaging the values - fixed: statsd metrics with non-ASCII text were processed by the statsd server, but were breaking JSON data generated by netdata - Add download_files _service - update source tarball to match upstream copy - Update to v1.7.0 * netdata is now a fully featured statsd server * improved metrics streaming and replication * backend enhancements, including: - host tags and metrics filtering on the netdata side - rewritten prometheus support to utilize more prometheus features and provide more flexibility and integration options * now monitors ZFS, ElasticSearch, RabbitMQ, Go applications (via expvar), samba, squid logs (with web_log plugin). * improved dashboard loading times * custom hook support for alarms - Update netdata-smartd-log-path.patch - mention patches added in last commit * add netdata-automake-no-dist-xz.patch * add netdata-logrotate-su.patch * add netdata-smartd-log-path.patch - Adjust default smartd log path to openSUSE default - Update dependencies * Enable more features * Add soft dependencies - Build on all enabled SUSE flavors - Improve package description - Use upstream .service/init.d/logrotate files (virtually identical) - Fix permissions - Clean up %install section (more readable) - spec-cleaner - Do not suppress errors from user/group creation. - Trim sensationalist wording from description. Replace -exec rm by just -delete. - Add netdata-rpmlintrc for spec file. - Fix url and source link. - Update to version 1.6.0 - Add systemd unit, init script and logrotate config from netdata git repo - Update to version 1.5.0 (see included ChangeLog) - Added netdata-rpmlintrc - Spec cleanup - Upgraded to 1.2.0 - Initial package, version: 1.0.1
Comments 0