Requirement
1.0 | Expose metrics from virtualized guests on a host system |
|
2.0 | Use libvirt API |
|
3.0 | Extend list of existing virt collectd plugin metrics for RDT monitoring |
|
4.0 | Should have a configurable interval |
|
5.0 | Provide SNMP support for added metrics |
|
Overview
Libvirt is an open source project licenced under GNU LGPL, which aims to provide a simple and convenient way of managing virtual machines created by various different hypervisors in a unified manner. It is an abstraction layer that provides a common application programming interface (API) for numerous functionalities implemented by hypervisors. Even if particular functionality is at the moment implemented only by a single hypervisor, it has to be exposed in a generic way to allow to support multiple hypervisors in the future if they eventually choose to implement the same feature. It makes learning hypervisor-specific tools no longer necessary, which is very advantageous and desirable characteristic in cloud-based environments. Major libvirt features include:
- Virtual machine management – domain lifecycle management, device hotplug operations
- Remote machine support – support for multiple network transports for remote connection
- Storage management – creation of file images, mounting NFS shares
- Network management – physical and logical network interfaces management
- Virtual NAT and route based networking – management and creation of virtual networks
Libvirt consists of 3 software components:
- API library
- Daemon - libvirtd
- Command line utility – virsh
Libvirt is based on client-server architecture, which requires daemon to be installed only on machine that will host virtualized guests. For the purpose of controlling domains remotely libvirt is using a custom protocol to establish connection with remote libvirtd instance. Although, it should be noted that domains can also be controlled locally, that is, client and server can be run on the same physical host.
Design
virt plugin
The virt plugin collects statistics by using virtualization API. Metrics are gathered directly from the hypervisor on a host system, which means that collectd doesn’t have to be installed and configured on a guest system. Plugin has been extended to support following metrics:
Type | Type Instance | Description | Comment |
cpu_affinity | vcpu_NR-cpu_NR | Pinning of domain VCPUs to host physical CPUs. | Value stored is a boolean. |
job_stats | * | Information about progress of a background/completed job on a domain. | Number of metrics depend on job type. Check API documentation for more information: virDomainGetJobStats |
disk_error | DISK_NAME | Disk error code | Metric is not dispatched for disk with no errors |
percent | virt_cpu_total | CPU utilization in percentage per domain | Computing percentage CPU utilization requires 2 samples of CPU time used by domain. Therefore, this metric will not be available on first dispatch of metrics after loading plugin or on first dispatch after domain restart operation. |
perf | * | Performance metrics | Number of metrics depend on libvirt API version. List of available perf metric can be found at: libvirt.org |
Additionally, support for following events has been added:
Type | Severity | Description | Comment |
domain_state | OKAY:
WARNING:
FAILURE:
| Domain state and reason in a human-readable format. |
|
file_system | OKAY | File system information:
| Information stored in metadata. Requires Guest Agent to be installed and configured in VM. |
Note:
Certain metrics and events have a requirement on a minimal libvirt API version. For more information, please see collectd virt plugin documentation.
Plugin configuration
Extended metric can be enabled with virt plugin configuration option:
Name | Selectors | Description | Comment |
ExtraStats |
| Defines whether additional statistics should be collected. By default no extra statistics are gathered, preserving the previous behavior of the plugin. | The argument is a space-separated list of selectors |
Here is an example of the plugin configuration section of collectd.conf file:
<Plugin virt>
RefreshInterval 60
ExtraStats "cpu_util disk_err domain_state fs_info job_stats_background perf vcpupin"
</Plugin>
Note:
Detailed information about plugin configuration options can be found in collectd virt plugin documentation.
SNMP Support
libvirt-snmp, libvirt subproject, offers only limited SNMP functionality, which doesn’t cover extended metrics enabled by SA project. If possible, existing MIB should be used for basic set of metrics and new MIB should be defined for extended metrics. Otherwise, new libvirt MIB should contain all metric gathered by virt collectd plugin. SNMP support can be achieved by adding proper configuration for snmp_agent collectd plugin. See description of SNMP feature for more details on snmp_agent plugin.
Considerations
Configuration Considerations
Deployment Considerations
It’s not recommended to enable extended perf metrics in conjunction with the Intel RDT plugin as both plugins are using same system resources.
Perf metric have to be supported by the platform.
API/GUI/CLI Considerations
Equivalence Considerations
The SNMP support for extended metric will be enabled by newly defined MIB.
Security Considerations
Alarms, events, statistics considerations
Certain hypervisors might not support all the metrics collected by the plugin. Unsupported metrics will not be reported.
Redundancy Considerations
Performance Considerations
Testing Consideration
Other Considerations
Impact
The following table outlines possible impact(s) the deployment of this deliverable may have on the current system.
Ref | System Impact Description | Recommendation / Comments |
1 |
|
|
Key Assumptions
The following assumptions apply to the scope specified in this document.
Ref | Assumption | Status |
1 |
|
|
Key Exclusions
The following exclusions apply to the scope discussed in this document.
Ref | Exclusion | Status |
1 |
|
|
Key Dependencies
The following table outlines the key dependencies associated with this deliverable.
Ref | Dependency | Status |
1 | libvirt |
|
2 | libxml2 |
|
3 |
|
|
4 |
|
|
Issues List
Ref | Issue | Status |
1 |
|
|