Anuket Project

Barometer Development Updates

Collectd plugin development status

PluginStatusDescriptionCommentsOPNFV ReleaseOpenstack Release
HugepagesupstreamedPlugin provides the free and used hugepage numbers/bytes/percentageUseful for intelligent workload placement for VMs backed by hugepagesDP
CacheupstreamedPlugin provides the last level cache utilitzation and memory bw utilizationBased on Resource Directory TechnologyEP
DPDK stats

upstreamed

Plugin provides the extended NIC stats for DPDK interfaces DP
DPDK eventsupstreamedPlugin provides the packet processing core status and the link status for DPDK interfaces EP
RAS Memory

upstreamed

Plugin uses mcelog client protocol to check for memory Machine Check Exceptions and sends the stats for reported exceptions. DP
BIOSReworked as a utility, waiting on snmp write plugin to be upstreamed  EQ
Open vSwitch Stats

upstreamed

Plugin provides the OVS stats for interfaces. Plugin is DPDK agnostic and uses the OVS DB. EP
Open vSwitch Events

upstreamed

Plugin provides the OVS link status for interfaces. Plugin will also report vswitch liveliness. DP
SNMP writeupstreamedPlugin will act as a SNMP subagent and will map collectd metrics to relavent OIDs. Will only support SNMP: get, getnext and walk. EQ
Legacy/IPMI

upstreamed

Plugin will report platform thermals, voltages, fanspeed.... EQ
RAS other errors

Implementation

Parsing and filtering utility for logfile implemented https://github.com/collectd/collectd/pull/2154 - reviewed, reworked and pending further review

Plugin will parse the mcelog/syslog for exceptions that are not memory exceptions. EQ
Libvirt extensionsUpstreamedExtend the libvirt plugin to include all relavent stats and events that are available for a libvirt Domain EQ
Python Notification

upstreamed

Extend the python language binding to pass the collectd metadata to write/notification plugins. EQ
PMUupstreamedPlugin will retirieve performance monitoring units (PMUs) that allow to count and sample a wide variety of events. EQ
PCIe AERhttps://github.com/maryamtahhan/collectd/tree/feat_pcie_errorsPlugin will poll PCI config space for baseline and AER errors. It will also parse syslog for AER events. Any errors to be reported via notifications. EQ
VES Collectd Agent update schema 5.1Merged to Batometer Repo  E 
OVS events Make the polling option configurableupstreamed  E 
mcelog updates

add a persist option for events - implemented - Pending Manual Test

send corrected and uncorrected errors in separate notifications with different severity

Merged to master

  E 
dpdkstat bugfixingEnable all ports - default config - Done  E 
Collectd extension

Internal protocol Extension - PUT to send commands to teh daemon

New plugin registration function

Dynamic Plugin - Load/unload

Internal code review completed, manual testing to be scheduled prior to upstreaming

 

  E 
QATLooking to leverage existing PCIe AER plugin to report QAT errorsError reporting from Accelerator E 
Collectd RPM buildMergedWill provide a base for the Apex installer work E 
OpenStack Boston DemoCompleteNoisy Neighbour detection a collaboration with Intel RDT and Vitrage   
Libvirt Plugin bugfixesIn progress

Domain restart causing incorrect CPU utilization Values to be reported

inconsistency in memory metrics reported by libvirt and virsh

   
DPDK events bugfixesMerged to collectd master

Reconnect to DPDK primary

wrong plugin config issues

respawn secondary after primary restart

   
Container /Cadvisor Support

Looking at integrating with Prometheus.

Investigating events support.

    
Process MonitoringDesign ReadyAdd a method which will accept a userspace process name as a parameter and return its pid   
OVS with DPDK PMD statsMerged to barometer repoCreating a script that will pull OVS-with-dpdk PMD stats and publish them to collectd. these stats are not retrievable through the OVSDB - but should be in the future - which is why this integration is stand alone for now  F
OVS Multi Instance Support

Pull request created:

https://github.com/maryamtahhan/collectd/tree/feat_ovs_multi_instance

OvS DB multi-instance support for ovs_event/ovs_stats plugins   
CI buildautomated, RPMs generated nightly and pushed to http://artifacts.opnfv.org/barometer.html   
Add PMU build to src/Under Review    
Create a docker image for barometerComplete  
Integrate with Apex

Integrated with 2 scenarios
apex-os-nosdn-bar-ha-baremetal-master

apex-os-nosdn-bar-noha-baremetal-master

   F Q
VES Kafka integrationSeparate out VES collectd application from collectd so that it can be apache licensed, and use a kafka bus to send info from collectd to this VES collector...    F
VES configuration through a schemaUnder internal review    
Connectivity. Event-based, interface monitorPull Request created: https://github.com/collectd/collectd/pull/2407This plugin monitors an Ethernet interface or group of ethernet interfaces and reports link status changes. The plugin does not poll for link status but rather recieves status events eliminating the polling overhead. Detection times vary by device driver, but typical detection times are below 10ms.  F Q
Procevent. Event-based process monitor.https://github.com/collectd/collectd/pull/2623This plugin monitors netlink process events and looks for status changes.
FQ
Sysevent. Event-based syslog monitorhttps://github.com/collectd/collectd/pull/2624This plugin monitors syslog and uses regex's to determine when to sent an event
FQ

 

Open Pull Requests