1 .. This work is licensed under a Creative Commons Attribution 4.0 International License.
2 .. http://creativecommons.org/licenses/by/4.0
3 .. (c) Anuket and others
5 ===========================
6 Anuket Barometer User Guide
7 ===========================
9 Barometer collectd plugins description
10 ---------------------------------------
11 .. Describe the specific features and how it is realised in the scenario in a brief manner
12 .. to ensure the user understand the context for the user guide instructions to follow.
14 Collectd is a daemon which collects system performance statistics periodically
15 and provides a variety of mechanisms to publish the collected metrics. It
16 supports more than 90 different input and output plugins. Input plugins
17 retrieve metrics and publish them to the collectd deamon, while output plugins
18 publish the data they receive to an end point. collectd also has infrastructure
19 to support thresholding and notification.
21 Barometer has enabled the following collectd plugins:
23 * *dpdkstat plugin*: A read plugin that retrieves stats from the DPDK extended
26 * *dpdkevents plugin*: A read plugin that retrieves DPDK link status and DPDK
27 forwarding cores liveliness status (DPDK Keep Alive).
29 * *dpdk_telemetry plugin*: A read plugin to collect dpdk interface stats and
30 application or global stats from dpdk telemetry library. Both 'dpdkstat' and
31 'dpdk_telemetry' plugins provides dpdk NIC Stats, but only 'dpdk_telemetry'
32 provides the DPDK Application stats. So in other words, 'dpdk_telemetry' is
33 an advanced version of dpdkstat. This plugin don't deal with dpdk events.
34 So not in related with 'dpdkevents' plugin. The mimimum dpdk version required
35 to use this plugin is 19.08.
38 dpdpkstat and dpdk_telemetry should not be used together. Use dpdk_telemetry
39 if your version of dpdk supports it (i.e. DPDK >= 19.08) and use dpdkstat otherwise.
40 dpdkstat, dpdkevents and dpdk_telemetry plugins should only be used if your dpdk
41 application doesn't already have more relevant metrics available(e.g.ovs_stats).
44 * `gnocchi plugin`_: A write plugin that pushes the retrieved stats to
45 Gnocchi. It's capable of pushing any stats read through collectd to
46 Gnocchi, not just the DPDK stats.
48 * `aodh plugin`_: A notification plugin that pushes events to Aodh, and
49 creates/updates alarms appropriately.
51 * *hugepages plugin*: A read plugin that retrieves the number of available
52 and free hugepages on a platform as well as what is available in terms of
55 * *Open vSwitch events Plugin*: A read plugin that retrieves events from OVS.
57 * *Open vSwitch stats Plugin*: A read plugin that retrieves flow and interface
60 * *mcelog plugin*: A read plugin that uses mcelog client protocol to check for
61 memory Machine Check Exceptions and sends the stats for reported exceptions.
63 * *PMU plugin*: A read plugin that provides performance counters data on
64 Intel CPUs using Linux perf interface.
66 * *RDT plugin*: A read plugin that provides the last level cache utilization and
67 memory bandwidth utilization.
69 * *virt*: A read plugin that uses virtualization API *libvirt* to gather
70 statistics about virtualized guests on a system directly from the hypervisor,
71 without a need to install collectd instance on the guest.
73 * *SNMP Agent*: A write plugin that will act as a AgentX subagent that receives
74 and handles queries from SNMP master agent and returns the data collected
75 by read plugins. The SNMP Agent plugin handles requests only for OIDs
76 specified in configuration file. To handle SNMP queries the plugin gets data
77 from collectd and translates requested values from collectd's internal format
78 to SNMP format. Supports SNMP: get, getnext and walk requests.
80 All the plugins above are available on the collectd main branch, except for
81 the Gnocchi and Aodh plugins as they are Python-based plugins and only C
82 plugins are accepted by the collectd community. The Gnocchi and Aodh plugins
83 live in the OpenStack repositories.
85 .. TODO: Update this to reflect merging of these PRs
86 Other plugins existing as a pull request into collectd main:
88 * *Legacy/IPMI*: A read plugin that reports platform thermals, voltages,
89 fanspeed, current, flow, power etc. Also, the plugin monitors Intelligent
90 Platform Management Interface (IPMI) System Event Log (SEL) and sends the
91 appropriate notifications based on monitored SEL events.
93 * *PCIe AER*: A read plugin that monitors PCIe standard and advanced errors and
94 sends notifications about those errors.
97 Third party application in Barometer repository:
99 * *Open vSwitch PMD stats*: An aplication that retrieves PMD stats from OVS. It is run
102 **Plugins and application included in the Euphrates release:**
104 Write Plugins: aodh plugin, SNMP agent plugin, gnocchi plugin.
106 Read Plugins/application: Intel RDT plugin, virt plugin, Open vSwitch stats plugin,
107 Open vSwitch PMD stats application.
109 Collectd capabilities and usage
110 ------------------------------------
111 .. Describe the specific capabilities and usage for <XYZ> feature.
112 .. Provide enough information that a user will be able to operate the feature on a deployed scenario.
114 The collectd plugins in Anuket are configured with reasonable defaults, but can
117 Building all Barometer upstreamed plugins from scratch
118 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
119 The plugins that have been merged to the collectd main branch can all be
120 built and configured through the barometer repository.
123 * sudo permissions are required to install collectd.
124 * These instructions are for Centos 7.
126 To build all the upstream plugins, clone the barometer repo:
130 $ git clone https://gerrit.opnfv.org/gerrit/barometer
132 To install collectd as a service and install all it's dependencies:
136 $ cd barometer/systems && ./build_base_machine.sh
138 This will install collectd as a service and the base install directory
139 will be /opt/collectd.
141 Sample configuration files can be found in '/opt/collectd/etc/collectd.conf.d'
144 If you don't want to use one of the Barometer plugins, simply remove the
145 sample config file from '/opt/collectd/etc/collectd.conf.d'
147 If you plan on using the Exec plugin (for OVS_PMD_STATS or for executing scripts
148 on notification generation), the plugin requires a non-root user to execute scripts.
149 By default, `collectd_exec` user is used in the exec.conf provided in the sample
150 configurations directory under src/collectd in the Barometer repo. These scripts *DO NOT* create this user.
151 You need to create this user or modify the configuration in the sample configurations directory
152 under src/collectd to use another existing non root user before running build_base_machine.sh.
155 If you are using any Open vSwitch plugins you need to run:
159 $ sudo ovs-vsctl set-manager ptcp:6640
161 After this, you should be able to start collectd as a service
165 $ sudo systemctl status collectd
167 If you want to use granfana to display the metrics you collect, please see:
170 For more information on configuring and installing OpenStack plugins for
171 collectd, check out the `collectd-openstack-plugins GSG`_.
173 Below is the per plugin installation and configuration guide, if you only want
174 to install some/particular plugins.
178 Repo: https://github.com/collectd/collectd
182 Dependencies: DPDK (https://dpdk.org/)
184 .. note:: DPDK statistics plugin requires DPDK version 16.04 or later.
186 To build and install DPDK to /usr please see:
187 https://github.com/collectd/collectd/blob/main/docs/BUILD.dpdkstat.md
189 Building and installing collectd:
193 $ git clone https://github.com/collectd/collectd.git
196 $ ./configure --enable-syslog --enable-logfile --enable-debug
200 .. note:: If DPDK was installed in a non standard location you will need to
201 specify paths to the header files and libraries using *LIBDPDK_CPPFLAGS* and
202 *LIBDPDK_LDFLAGS*. You will also need to add the DPDK library symbols to the
203 shared library path using *ldconfig*. Note that this update to the shared
204 library path is not persistant (i.e. it will not survive a reboot).
206 Example of specifying custom paths to DPDK headers and libraries:
210 $ ./configure LIBDPDK_CPPFLAGS="path to DPDK header files" LIBDPDK_LDFLAGS="path to DPDK libraries"
212 This will install collectd to default folder ``/opt/collectd``. The collectd
213 configuration file (``collectd.conf``) can be found at ``/opt/collectd/etc``.
214 To configure the dpdkstats plugin you need to modify the configuration file to
222 ProcessType "secondary"
224 EnabledPortMask 0xffff
225 PortName "interface1"
226 PortName "interface2"
230 To configure the dpdkevents plugin you need to modify the configuration file to
235 <LoadPlugin dpdkevents>
239 <Plugin "dpdkevents">
245 <Event "link_status">
246 SendEventsOnUpdate false
247 EnabledPortMask 0xffff
248 SendNotification true
251 SendEventsOnUpdate false
253 KeepAliveShmName "/dpdk_keepalive_shm_name"
254 SendNotification true
258 .. note:: Currently, the DPDK library doesn’t support API to de-initialize
259 the DPDK resources allocated on the initialization. It means, the collectd
260 plugin will not be able to release the allocated DPDK resources
261 (locks/memory/pci bindings etc.) correctly on collectd shutdown or reinitialize
262 the DPDK library if primary DPDK process is restarted. The only way to release
263 those resources is to terminate the process itself. For this reason, the plugin
264 forks off a separate collectd process. This child process becomes a secondary
265 DPDK process which can be run on specific CPU cores configured by user through
266 collectd configuration file (“Coremask” EAL configuration option, the
267 hexadecimal bitmask of the cores to run on).
269 For more information on the plugin parameters, please see:
270 https://github.com/collectd/collectd/blob/main/src/collectd.conf.pod
272 .. note:: dpdkstat plugin initialization time depends on read interval. It
273 requires 5 read cycles to set up internal buffers and states, during that time
274 no statistics are submitted. Also, if plugin is running and the number of DPDK
275 ports is increased, internal buffers are resized. That requires 3 read cycles
276 and no port statistics are submitted during that time.
278 The Address-Space Layout Randomization (ASLR) security feature in Linux should be
279 disabled, in order for the same hugepage memory mappings to be present in all
280 DPDK multi-process applications.
286 $ sudo echo 0 > /proc/sys/kernel/randomize_va_space
288 To fully enable ASLR:
292 $ sudo echo 2 > /proc/sys/kernel/randomize_va_space
294 .. warning:: Disabling Address-Space Layout Randomization (ASLR) may have security
295 implications. It is recommended to be disabled only when absolutely necessary,
296 and only when all implications of this change have been understood.
298 For more information on multi-process support, please see:
299 https://doc.dpdk.org/guides/prog_guide/multi_proc_support.html
301 **DPDK stats plugin limitations:**
303 1. The DPDK primary process application should use the same version of DPDK
304 that collectd DPDK plugin is using;
306 2. L2 statistics are only supported;
308 3. The plugin has been tested on Intel NIC’s only.
310 **DPDK stats known issues:**
312 * DPDK port visibility
314 When network port controlled by Linux is bound to DPDK driver, the port
315 will not be available in the OS. It affects the SNMP write plugin as those
316 ports will not be present in standard IF-MIB. Thus, additional work is
317 required to be done to support DPDK ports and statistics.
319 **DPDK telemetry plugin**
321 Please refer to https://wiki.anuket.io/display/HOME/DPDK+Telemetry+Plugin
325 Repo: https://github.com/collectd/collectd
329 Dependencies: None, but assumes hugepages are configured.
331 To configure some hugepages:
335 $ sudo mkdir -p /mnt/huge
336 $ sudo mount -t hugetlbfs nodev /mnt/huge
337 $ sudo bash -c "echo 14336 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages"
339 Building and installing collectd:
343 $ git clone https://github.com/collectd/collectd.git
346 $ ./configure --enable-syslog --enable-logfile --enable-hugepages --enable-debug
350 This will install collectd to default folder ``/opt/collectd``. The collectd
351 configuration file (``collectd.conf``) can be found at ``/opt/collectd/etc``.
352 To configure the hugepages plugin you need to modify the configuration file to
363 ValuesPercentage false
366 For more information on the plugin parameters, please see:
367 https://github.com/collectd/collectd/blob/main/src/collectd.conf.pod
371 Repo: https://github.com/collectd/collectd
377 * PMU tools (jevents library) https://github.com/andikleen/pmu-tools
379 To be suitable for use in collectd plugin shared library *libjevents* should be
380 compiled as position-independent code. To do this add the following line to
381 *pmu-tools/jevents/Makefile*:
387 Building and installing *jevents* library:
391 $ git clone https://github.com/andikleen/pmu-tools.git
392 $ cd pmu-tools/jevents/
396 Download the Hardware Events that are relevant to your CPU, download the appropriate
397 CPU event list json file:
401 $ wget https://raw.githubusercontent.com/andikleen/pmu-tools/main/event_download.py
402 $ python event_download.py
404 This will download the json files to the location: $HOME/.cache/pmu-events/. If you don't want to
405 download these files to the aforementioned location, set the environment variable XDG_CACHE_HOME to
406 the location you want the files downloaded to.
408 Building and installing collectd:
412 $ git clone https://github.com/collectd/collectd.git
415 $ ./configure --enable-syslog --enable-logfile --with-libjevents=/usr/local --enable-debug
419 This will install collectd to default folder ``/opt/collectd``. The collectd
420 configuration file (``collectd.conf``) can be found at ``/opt/collectd/etc``.
421 To configure the PMU plugin you need to modify the configuration file to
426 <LoadPlugin intel_pmu>
430 ReportHardwareCacheEvents true
431 ReportKernelPMUEvents true
432 ReportSoftwareEvents true
436 If you want to monitor Intel CPU specific CPU events, make sure to enable the
437 additional two options shown below:
442 ReportHardwareCacheEvents true
443 ReportKernelPMUEvents true
444 ReportSoftwareEvents true
445 EventList "$HOME/.cache/pmu-events/GenuineIntel-6-2D-core.json"
446 HardwareEvents "L2_RQSTS.CODE_RD_HIT,L2_RQSTS.CODE_RD_MISS" "L2_RQSTS.ALL_CODE_RD"
451 If you set XDG_CACHE_HOME to anything other than the variable above - you will need to modify
452 the path for the EventList configuration.
454 Use "Cores" option to monitor metrics only for configured cores. If an empty string is provided
455 as value for this field default cores configuration is applied - that is all available cores
456 are monitored separately. To limit monitoring to cores 0-7 set the option as shown below:
462 For more information on the plugin parameters, please see:
463 https://github.com/collectd/collectd/blob/main/src/collectd.conf.pod
467 The plugin opens file descriptors whose quantity depends on number of
468 monitored CPUs and number of monitored counters. Depending on configuration,
469 it might be required to increase the limit on the number of open file
470 descriptors allowed. This can be done using 'ulimit -n' command. If collectd
471 is executed as a service 'LimitNOFILE=' directive should be defined in
472 [Service] section of *collectd.service* file.
476 Repo: https://github.com/collectd/collectd
482 * PQoS/Intel RDT library https://github.com/intel/intel-cmt-cat
485 Building and installing PQoS/Intel RDT library:
489 $ git clone https://github.com/intel/intel-cmt-cat
492 $ make install PREFIX=/usr
494 You will need to insert the msr kernel module:
500 Building and installing collectd:
504 $ git clone https://github.com/collectd/collectd.git
507 $ ./configure --enable-syslog --enable-logfile --with-libpqos=/usr/ --enable-debug
511 This will install collectd to default folder ``/opt/collectd``. The collectd
512 configuration file (``collectd.conf``) can be found at ``/opt/collectd/etc``.
513 To configure the RDT plugin you need to modify the configuration file to
518 <LoadPlugin intel_rdt>
525 For more information on the plugin parameters, please see:
526 https://github.com/collectd/collectd/blob/main/src/collectd.conf.pod
530 Repo: https://github.com/collectd/collectd
532 Branch: feat_ipmi_events, feat_ipmi_analog
534 Dependencies: OpenIPMI library (http://openipmi.sourceforge.net/)
536 The IPMI plugin is already implemented in the latest collectd and sensors
537 like temperature, voltage, fanspeed, current are already supported there.
538 The list of supported IPMI sensors has been extended and sensors like flow,
539 power are supported now. Also, a System Event Log (SEL) notification feature
542 * The feat_ipmi_events branch includes new SEL feature support in collectd
543 IPMI plugin. If this feature is enabled, the collectd IPMI plugin will
544 dispatch notifications about new events in System Event Log.
546 * The feat_ipmi_analog branch includes the support of extended IPMI sensors in
547 collectd IPMI plugin.
549 **Install dependencies**
551 On Centos, install OpenIPMI library:
555 $ sudo yum install OpenIPMI ipmitool
557 Anyway, it's recommended to use the latest version of the OpenIPMI library as
558 it includes fixes of known issues which aren't included in standard OpenIPMI
559 library package. The latest version of the library can be found at
560 https://sourceforge.net/p/openipmi/code/ci/master/tree/. Steps to install the
561 library from sources are described below.
563 Remove old version of OpenIPMI library:
567 $ sudo yum remove OpenIPMI ipmitool
569 Build and install OpenIPMI library:
573 $ git clone https://git.code.sf.net/p/openipmi/code openipmi-code
575 $ autoreconf --install
576 $ ./configure --prefix=/usr
580 Add the directory containing ``OpenIPMI*.pc`` files to the ``PKG_CONFIG_PATH``
581 environment variable:
585 export PKG_CONFIG_PATH=/usr/lib/pkgconfig
587 Enable IPMI support in the kernel:
591 $ sudo modprobe ipmi_devintf
592 $ sudo modprobe ipmi_si
595 If HW supports IPMI, the ``/dev/ipmi0`` character device will be
598 Clone and install the collectd IPMI plugin:
602 $ git clone https://github.com/collectd/collectd
605 $ ./configure --enable-syslog --enable-logfile --enable-debug
609 This will install collectd to default folder ``/opt/collectd``. The collectd
610 configuration file (``collectd.conf``) can be found at ``/opt/collectd/etc``.
611 To configure the IPMI plugin you need to modify the file to include:
618 SELEnabled true # only feat_ipmi_events branch supports this
623 By default, IPMI plugin will read all available analog sensor values,
624 dispatch the values to collectd and send SEL notifications.
626 For more information on the IPMI plugin parameters and SEL feature configuration,
627 please see: https://github.com/collectd/collectd/blob/main/src/collectd.conf.pod
629 Extended analog sensors support doesn't require additional configuration. The usual
630 collectd IPMI documentation can be used:
632 - https://collectd.org/wiki/index.php/Plugin:IPMI
633 - https://collectd.org/documentation/manpages/collectd.conf.5.shtml#plugin_ipmi
637 - https://www.kernel.org/doc/Documentation/IPMI.txt
638 - https://www.intel.com/content/www/us/en/products/docs/servers/ipmi/ipmi-second-gen-interface-spec-v2-rev1-1.html
642 Repo: https://github.com/collectd/collectd
648 Start by installing mcelog.
651 The kernel has to have CONFIG_X86_MCE enabled. For 32bit kernels you
652 need atleast a 2.6,30 kernel.
658 $ sudo yum install mcelog
664 $ git clone https://git.kernel.org/pub/scm/utils/cpu/mce/mcelog.git
669 $ cp mcelog.service /etc/systemd/system/
670 $ systemctl enable mcelog.service
671 $ systemctl start mcelog.service
674 Verify you got a /dev/mcelog. You can verify the daemon is running completely
681 This should query the information in the running daemon. If it prints nothing
682 that is fine (no errors logged yet). More info @
683 http://www.mcelog.org/installation.html
685 Modify the mcelog configuration file "/etc/mcelog/mcelog.conf" to include or
690 socket-path = /var/run/mcelog-client
692 dimm-tracking-enabled = yes
693 dmi-prepopulate = yes
694 uc-error-threshold = 1 / 24h
695 ce-error-threshold = 10 / 24h
698 socket-tracking-enabled = yes
699 mem-uc-error-threshold = 100 / 24h
700 mem-ce-error-threshold = 100 / 24h
701 mem-ce-error-log = yes
704 memory-ce-threshold = 10 / 24h
706 memory-ce-action = soft
710 directory = /etc/mcelog
713 Clone and install the collectd mcelog plugin:
717 $ git clone https://github.com/collectd/collectd
720 $ ./configure --enable-syslog --enable-logfile --enable-debug
724 This will install collectd to default folder ``/opt/collectd``. The collectd
725 configuration file (``collectd.conf``) can be found at ``/opt/collectd/etc``.
726 To configure the mcelog plugin you need to modify the configuration file to
736 McelogClientSocket "/var/run/mcelog-client"
737 PersistentNotification false
739 #McelogLogfile "/var/log/mcelog"
742 For more information on the plugin parameters, please see:
743 https://github.com/collectd/collectd/blob/main/src/collectd.conf.pod
745 Simulating a Machine Check Exception can be done in one of 3 ways:
747 * Running $make test in the mcelog cloned directory - mcelog test suite
751 **mcelog test suite:**
753 It is always a good idea to test an error handling mechanism before it is
754 really needed. mcelog includes a test suite. The test suite relies on
755 mce-inject which needs to be installed and in $PATH.
757 You also need the mce-inject kernel module configured (with
758 CONFIG_X86_MCE_INJECT=y), compiled, installed and loaded:
762 $ modprobe mce-inject
764 Then you can run the mcelog test suite with
770 This will inject different classes of errors and check that the mcelog triggers
771 runs. There will be some kernel messages about page offlining attempts. The
772 test will also lose a few pages of memory in your system (not significant).
775 This test will kill any running mcelog, which needs to be restarted
780 A utility to inject corrected, uncorrected and fatal machine check exceptions
784 $ git clone https://git.kernel.org/pub/scm/utils/cpu/mce/mce-inject.git
787 $ modprobe mce-inject
789 Modify the test/corrected script to include the following:
794 STATUS 0xcc00008000010090
800 $ ./mce-inject < test/corrected
803 The uncorrected and fatal scripts under test will cause a platform reset.
804 Only the fatal script generates the memory errors**. In order to quickly
805 emulate uncorrected memory errors and avoid host reboot following test errors
806 from mce-test suite can be injected:
810 $ mce-inject mce-test/cases/coverage/soft-inj/recoverable_ucr/data/srao_mem_scrub
814 In addition a more in-depth test of the Linux kernel machine check facilities
815 can be done with the mce-test test suite. mce-test supports testing uncorrected
816 error handling, real error injection, handling of different soft offlining
817 cases, and other tests.
819 **Corrected memory error injection:**
821 To inject corrected memory errors:
823 * Remove sb_edac and edac_core kernel modules: rmmod sb_edac rmmod edac_core
824 * Insert einj module: modprobe einj param_extension=1
825 * Inject an error by specifying details (last command should be repeated at least two times):
829 $ APEI_IF=/sys/kernel/debug/apei/einj
830 $ echo 0x8 > $APEI_IF/error_type
831 $ echo 0x01f5591000 > $APEI_IF/param1
832 $ echo 0xfffffffffffff000 > $APEI_IF/param2
833 $ echo 1 > $APEI_IF/notrigger
834 $ echo 1 > $APEI_IF/error_inject
836 * Check the MCE statistic: mcelog --client. Check the mcelog log for injected error details: less /var/log/mcelog.
839 ^^^^^^^^^^^^^^^^^^^^^
840 OvS Plugins Repo: https://github.com/collectd/collectd
842 OvS Plugins Branch: main
844 OvS Events MIBs: The SNMP OVS interface link status is provided by standard
845 IF-MIB (http://www.net-snmp.org/docs/mibs/IF-MIB.txt)
847 Dependencies: Open vSwitch, Yet Another JSON Library (https://github.com/lloyd/yajl)
849 On Centos, install the dependencies and Open vSwitch:
853 $ sudo yum install yajl-devel
855 Steps to install Open vSwtich can be found at
856 https://docs.openvswitch.org/en/latest/intro/install/fedora/
858 Start the Open vSwitch service:
862 $ sudo service openvswitch-switch start
864 Configure the ovsdb-server manager:
868 $ sudo ovs-vsctl set-manager ptcp:6640
870 Clone and install the collectd ovs plugin:
878 $ ./configure --enable-syslog --enable-logfile --enable-debug
882 This will install collectd to default folder ``/opt/collectd``. The collectd
883 configuration file (``collectd.conf``) can be found at ``/opt/collectd/etc``.
884 To configure the OVS events plugin you need to modify the configuration file to include:
888 <LoadPlugin ovs_events>
894 Socket "/var/run/openvswitch/db.sock"
895 Interfaces "br0" "veth0"
896 SendNotification true
899 To configure the OVS stats plugin you need to modify the configuration file
904 <LoadPlugin ovs_stats>
910 Socket "/var/run/openvswitch/db.sock"
914 For more information on the plugin parameters, please see:
915 https://github.com/collectd/collectd/blob/main/src/collectd.conf.pod
919 Repo: https://gerrit.opnfv.org/gerrit/barometer
922 1. Open vSwitch dependencies are installed.
923 2. Open vSwitch service is running.
924 3. Ovsdb-server manager is configured.
925 You can refer `Open vSwitch Plugins`_ section above for each one of them.
927 OVS PMD stats application is run through the exec plugin.
929 To configure the OVS PMD stats application you need to modify the exec plugin configuration
938 Exec "user:group" "<path to ovs_pmd_stat.sh>"
941 .. note:: Exec plugin configuration has to be changed to use appropriate user before starting collectd service.
943 ovs_pmd_stat.sh calls the script for OVS PMD stats application with its argument:
947 sudo python /usr/local/src/ovs_pmd_stats.py" "--socket-pid-file"
948 "/var/run/openvswitch/ovs-vswitchd.pid"
952 Repo: https://github.com/collectd/collectd
956 Dependencies: NET-SNMP library
958 Start by installing net-snmp and dependencies.
964 $ sudo yum install net-snmp net-snmp-libs net-snmp-utils net-snmp-devel
965 $ sudo systemctl start snmpd.service
967 go to the `snmp configuration`_ steps.
971 Clone and build net-snmp:
975 $ git clone https://github.com/haad/net-snmp.git
977 $ ./configure --with-persistent-directory="/var/net-snmp" --with-systemd --enable-shared --prefix=/usr
986 Copy default configuration to persistent folder:
990 $ cp EXAMPLE.conf /usr/share/snmp/snmpd.conf
992 Set library path and default MIB configuration:
997 $ echo export LD_LIBRARY_PATH=/usr/lib >> .bashrc
998 $ net-snmp-config --default-mibdirs
999 $ net-snmp-config --snmpconfpath
1001 Configure snmpd as a service:
1006 $ cp ./dist/snmpd.service /etc/systemd/system/
1007 $ systemctl enable snmpd.service
1008 $ systemctl start snmpd.service
1010 .. _`snmp configuration`:
1012 Add the following line to snmpd.conf configuration file
1013 ``/etc/snmp/snmpd.conf`` to make all OID tree visible for SNMP clients:
1017 view systemview included .1
1019 To verify that SNMP is working you can get IF-MIB table using SNMP client
1020 to view the list of Linux interfaces:
1024 $ snmpwalk -v 2c -c public localhost IF-MIB::interfaces
1026 Get the default MIB location:
1030 $ net-snmp-config --default-mibdirs
1031 /opt/stack/.snmp/mibs:/usr/share/snmp/mibs
1033 Install Intel specific MIBs (if needed) into location received by
1034 ``net-snmp-config`` command (e.g. ``/usr/share/snmp/mibs``).
1038 $ git clone https://gerrit.opnfv.org/gerrit/barometer.git
1039 $ sudo cp -f barometer/mibs/*.txt /usr/share/snmp/mibs/
1040 $ sudo systemctl restart snmpd.service
1042 Clone and install the collectd snmp_agent plugin:
1047 $ git clone https://github.com/collectd/collectd
1050 $ ./configure --enable-syslog --enable-logfile --enable-debug --enable-snmp --with-libnetsnmp
1054 This will install collectd to default folder ``/opt/collectd``. The collectd
1055 configuration file (``collectd.conf``) can be found at ``/opt/collectd/etc``.
1057 **SNMP Agent plugin is a generic plugin and cannot work without configuration**.
1058 To configure the snmp_agent plugin you need to modify the configuration file to
1059 include OIDs mapped to collectd types. The following example maps scalar
1060 memAvailReal OID to value represented as free memory type of memory plugin:
1064 LoadPlugin snmp_agent
1065 <Plugin "snmp_agent">
1066 <Data "memAvailReal">
1070 OIDs "1.3.6.1.4.1.2021.4.6.0"
1075 The ``snmpwalk`` command can be used to validate the collectd configuration:
1079 $ snmpwalk -v 2c -c public localhost 1.3.6.1.4.1.2021.4.6.0
1080 UCD-SNMP-MIB::memAvailReal.0 = INTEGER: 135237632 kB
1085 * Object instance with Counter64 type is not supported in SNMPv1. When GetNext
1086 request is received, Counter64 type objects will be skipped. When Get
1087 request is received for Counter64 type object, the error will be returned.
1088 * Interfaces that are not visible to Linux like DPDK interfaces cannot be
1089 retreived using standard IF-MIB tables.
1091 For more information on the plugin parameters, please see:
1092 https://github.com/collectd/collectd/blob/main/src/collectd.conf.pod
1094 For more details on AgentX subagent, please see:
1095 http://www.net-snmp.org/tutorial/tutorial-5/toolkit/demon/
1101 Repo: https://github.com/collectd/collectd
1105 Dependencies: libvirt (https://libvirt.org/), libxml2
1107 On Centos, install the dependencies:
1111 $ sudo yum install libxml2-devel libpciaccess-devel yajl-devel device-mapper-devel
1115 .. note:: libvirt version in package manager might be quite old and offer only
1116 limited functionality. Hence, building and installing libvirt from sources
1117 is recommended. Detailed instructions can bet found at:
1118 https://libvirt.org/compiling.html
1122 $ sudo yum install libvirt-devel
1124 Certain metrics provided by the plugin have a requirement on a minimal version of
1125 the libvirt API. *File system information* statistics require a *Guest Agent (GA)*
1126 to be installed and configured in a VM. User must make sure that installed GA
1127 version supports retrieving file system information. Number of *Performance monitoring events*
1128 metrics depends on running libvirt daemon version.
1130 .. note:: Please keep in mind that RDT metrics (part of *Performance monitoring
1131 events*) have to be supported by hardware. For more details on hardware support,
1133 https://github.com/intel/intel-cmt-cat
1135 Additionally perf metrics **cannot** be collected if *Intel RDT* plugin is enabled.
1137 libvirt version can be checked with following commands:
1142 $ libvirtd --version
1144 .. table:: Extended statistics requirements
1146 +-------------------------------+--------------------------+-------------+
1147 | Statistic | Min. libvirt API version | Requires GA |
1148 +===============================+==========================+=============+
1149 | Domain reason | 0.9.2 | No |
1150 +-------------------------------+--------------------------+-------------+
1151 | Disk errors | 0.9.10 | No |
1152 +-------------------------------+--------------------------+-------------+
1153 | Job statistics | 1.2.9 | No |
1154 +-------------------------------+--------------------------+-------------+
1155 | File system information | 1.2.11 | Yes |
1156 +-------------------------------+--------------------------+-------------+
1157 | Performance monitoring events | 1.3.3 | No |
1158 +-------------------------------+--------------------------+-------------+
1160 Start libvirt daemon:
1164 $ systemctl start libvirtd
1166 Create domain (VM) XML configuration file. For more information on domain XML
1167 format and examples, please see:
1168 https://libvirt.org/formatdomain.html
1170 .. note:: Installing additional hypervisor dependencies might be required before
1171 deploying virtual machine.
1173 Create domain, based on created XML file:
1177 $ virsh define DOMAIN_CFG_FILE.xml
1183 $ virsh start DOMAIN_NAME
1185 Check if domain is running:
1191 Check list of available *Performance monitoring events* and their settings:
1195 $ virsh perf DOMAIN_NAME
1197 Enable or disable *Performance monitoring events* for domain:
1201 $ virsh perf DOMAIN_NAME [--enable | --disable] EVENT_NAME --live
1203 Clone and install the collectd virt plugin:
1210 $ ./configure --enable-syslog --enable-logfile --enable-debug
1214 Where ``$REPO`` is equal to information provided above.
1216 This will install collectd to ``/opt/collectd``. The collectd configuration file
1217 ``collectd.conf`` can be found at ``/opt/collectd/etc``.
1218 To load the virt plugin user needs to modify the configuration file to include:
1224 Additionally, user can specify plugin configuration parameters in this file,
1225 such as connection URL, domain name and much more. By default extended virt plugin
1226 statistics are disabled. They can be enabled with ``ExtraStats`` option.
1232 ExtraStats "cpu_util disk disk_err domain_state fs_info job_stats_background pcpu perf vcpupin"
1235 For more information on the plugin parameters, please see:
1236 https://github.com/collectd/collectd/blob/main/src/collectd.conf.pod
1238 .. _install-collectd-as-a-service:
1240 Installing collectd as a service
1241 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1243 Collectd service scripts are available in the collectd/contrib directory.
1244 To install collectd as a service:
1248 $ sudo cp contrib/systemd.collectd.service /etc/systemd/system/
1249 $ cd /etc/systemd/system/
1250 $ sudo mv systemd.collectd.service collectd.service
1251 $ sudo chmod +x collectd.service
1253 Modify collectd.service
1258 ExecStart=/opt/collectd/sbin/collectd
1259 EnvironmentFile=-/opt/collectd/etc/
1260 EnvironmentFile=-/opt/collectd/etc/
1261 CapabilityBoundingSet=CAP_SETUID CAP_SETGID
1267 $ sudo systemctl daemon-reload
1268 $ sudo systemctl start collectd.service
1269 $ sudo systemctl status collectd.service should show success
1271 Additional useful plugins
1272 ^^^^^^^^^^^^^^^^^^^^^^^^^^
1274 **Exec Plugin** : Can be used to show you when notifications are being
1275 generated by calling a bash script that dumps notifications to file. (handy
1276 for debug). Modify /opt/collectd/etc/collectd.conf:
1282 # Exec "user:group" "/path/to/exec"
1283 NotificationExec "user" "<path to barometer>/barometer/src/collectd/collectd_sample_configs/write_notification.sh"
1286 write_notification.sh (just writes the notification passed from exec through
1287 STDIN to a file (/tmp/notifications)):
1292 rm -f /tmp/notifications
1295 echo $x$y >> /tmp/notifications
1298 output to /tmp/notifications should look like:
1306 PluginInstance:br-ex
1308 TypeInstance:link_status
1309 uuid:f2aafeec-fa98-4e76-aec5-18ae9fc74589
1311 linkstate of "br-ex" interface has been changed to "DOWN"
1313 * **logfile plugin**: Can be used to log collectd activity. Modify
1314 /opt/collectd/etc/collectd.conf to include:
1321 File "/var/log/collectd.log"
1327 Monitoring Interfaces and Openstack Support
1328 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1329 .. Figure:: monitoring_interfaces.png
1331 Monitoring Interfaces and Openstack Support
1333 The figure above shows the DPDK L2 forwarding application running on a compute
1334 node, sending and receiving traffic. Collectd is also running on this compute
1335 node retrieving the stats periodically from DPDK through the dpdkstat plugin
1336 and publishing the retrieved stats to OpenStack through the
1337 collectd-openstack-plugins.
1339 To see this demo in action please checkout: `Barometer OPNFV Summit demo`_
1341 For more information on configuring and installing OpenStack plugins for
1342 collectd, check out the `collectd-openstack-plugins GSG`_.
1346 * AAA – on top of collectd there secure agents like SNMP V3, Openstack agents
1347 etc. with their own AAA methods.
1349 * Collectd runs as a daemon with root permissions.
1351 * The `Exec plugin`_ allows the execution of external programs but counters the security
1354 * Ensuring that only one instance of the program is executed by collectd at any time
1355 * Forcing the plugin to check that custom programs are never executed with superuser
1358 * Protection of Data in flight:
1360 * It's recommend to use a minimum version of 4.7 of the Network plugin which provides
1361 the possibility to cryptographically sign or encrypt the network traffic.
1362 * Write Redis plugin or the Write MongoDB plugin are recommended to store the data.
1363 * For more information, please see: https://collectd.org/wiki/index.php?title=Networking_introduction
1365 * Known vulnerabilities include:
1367 * https://www.cvedetails.com/vulnerability-list/vendor_id-11242/Collectd.html
1369 * `CVE-2017-7401`_ fixed https://github.com/collectd/collectd/issues/2174 in Version 5.7.2.
1370 * `CVE-2016-6254`_ fixed https://mailman.verplant.org/pipermail/collectd/2016-July/006838.html
1372 * `CVE-2010-4336`_ fixed https://mailman.verplant.org/pipermail/collectd/2010-November/004277.html
1375 * https://www.cvedetails.com/product/20310/Collectd-Collectd.html?vendor_id=11242
1377 * It's recommended to only use collectd plugins from signed packages.
1381 .. [1] https://collectd.org/wiki/index.php/Naming_schema
1382 .. [2] https://github.com/collectd/collectd/blob/main/src/daemon/plugin.h
1383 .. [3] https://collectd.org/wiki/index.php/Value_list_t
1384 .. [4] https://collectd.org/wiki/index.php/Data_set
1385 .. [5] https://collectd.org/documentation/manpages/types.db.5.shtml
1386 .. [6] https://collectd.org/wiki/index.php/Data_source
1387 .. [7] https://collectd.org/wiki/index.php/Meta_Data_Interface
1389 .. _Barometer OPNFV Summit demo: https://prezi.com/kjv6o8ixs6se/software-fastpath-service-quality-metrics-demo/
1390 .. _gnocchi plugin: https://github.com/openstack/collectd-openstack-plugins/tree/stable/ocata/
1391 .. _aodh plugin: https://github.com/openstack/collectd-openstack-plugins/tree/stable/ocata/
1392 .. _collectd-openstack-plugins GSG: https://opendev.org/x/collectd-openstack-plugins/src/branch/master/doc/source/GSG.rst
1393 .. _grafana guide: https://wiki.anuket.io/display/HOME/Installing+and+configuring+InfluxDB+and+Grafana+to+display+metrics+with+collectd
1394 .. _CVE-2017-7401: https://www.cvedetails.com/cve/CVE-2017-7401/
1395 .. _CVE-2016-6254: https://www.cvedetails.com/cve/CVE-2016-6254/
1396 .. _CVE-2010-4336: https://www.cvedetails.com/cve/CVE-2010-4336/
1397 .. _Exec plugin: https://collectd.org/wiki/index.php/Plugin:Exec