Download presentation
Published byCassandra Stone Modified over 8 years ago
2
Failure Inspection in Doctor utilizing Vitrage and Congress
Ryota Mibu, NEC Ohad Shamir, Nokia Masahito Muroi, NTT
3
Agenda Failure Inspection in OPNFV Doctor - Ryota OpenStack Vitrage - Ohad OpenStack Congress - Masa
4
Failure Inspection in OPNFV Doctor
5
Virtualized Platform VM VM VM Net Net PM PM H/W Switch H/W Switch
Volume Port Port Port Volume Net Net PM PM H/W Switch H/W Switch
6
Virtualized Platform X X VM VM VM Net Net PM PM H/W Switch H/W Switch
Volume X Port Port Port Volume Net Net PM PM H/W Switch H/W Switch
7
What is “failure”? Depends on So, “failure” has to be configurable
Applications (VNFs) Back-end technologies used in the deployment Redundancy of the equipment/components Operator Policy Regulation Topologies of Network / Power-supply So, “failure” has to be configurable
8
Fault Management Architecture
Designed by Doctor Project Applications Manager 0. Set Alarm 6-. Action 5. Notify Error Virtualized Infrastructure 4. Notify all Controller Controller Notifier Nova Controller Resource Map Alarm Conf. Neutron Ceilometer+Aodh Cinder 3. Update State 2. Find Affected 4. (alt) Notify Monitor Monitor Inspector Monitor Failure Policy Congress 1. Raw Fault Vitrage
9
Inspector Module Options
OpenStack Vitrage Various data source (OpenStack, Nagios, etc.) Ability to store and refer physical topologies for correlation Holistic & complete view of the system OpenStack Congress Dynamic data collection from OpenStack services Flexible policy definition for correlation (Datalog) Well integrated with other OpenStack projects
10
Stained glass art produced through the combination of brilliantly colored glass in varying degrees of transparency, creating a dynamic art form which is transformed with every variation in light Stained glass window Taken alone they don’t make sense Together they create a pattern
11
Vitrage in a nutshell Official OpenStack project for Root Cause Analysis Vitrage Functions Root Cause Analysis – understand what causes faults to occur Deduced alarms and states – raising alarms and modifying states based on system insights Holistic & complete view of the system Architecture Highlights Resource topology graph - reflect how different entities relate to one another Multiple data sources Configurable business logic Clear visualization of Vitrage insights
12
Vitrage High Level Architecture
Horizon Plug-in: Hierarchical view Vitrage alarm list RCA diagram per alarm Entity graph view Expose Vitrage alarms and state changes to other projects or external systems Multiple Data Sources (extendible): External monitoring tools: Nagios, Zabbix OpenStack projects Physical topology Templates for deduced alarms and RCA: Each template can contain one or more scenarios (scenario = condition + action) Human readable Configurable
13
How does it work? Entity Graph + Evaluator
Vitrage holds the system state (resources & alarms) as a property graph Entities are represented as graph vertices, and relationships are the edges between the vertices Each vertex and edge can also have additional properties Each edge has a special “label” indicating the type of relationship Intuitive modeling of relationship/interaction data Vitrage Evaluator listens to change events in the entity graph and upon event: Retrieve templates (scenarios) relevant to event Evaluate condition against the state of the Entity Graph Execute actions for each matched condition Switch Name: sw1 id: XXX Host Name: comp 0-0 attached Instance Status: Error contains Alarm on
14
Vitrage User Interface (Horizon plug-in)
15
Vitrage Example Host NIC failure (no HA)
Monitor host NIC (public/ tenant network) by Nagios/ Zabbix, raise an alarm when failed Vitrage will receive alarm from monitoring tool, add it to the entity graph, connected to the Host NIC vertex Find matching scenario (template) and perform the following actions: Raise deduce alarm on host add it to the vitrage entity graph as well Change host state in Vitrage may trigger also calling Nova API to modify state Add causal relationship between alarms Once the deduced alarm on host is added, a similar flow will occur on the hosted instances and on relevant VNF Zone 1 Alarm 1 causes on attached Host NIC Host 1 Host 2 on contains Alarm 2 causes on VM 1 Alarm 3 causes on Alarm 4 VNF
16
Vitrage as Doctor Inspector
Push and pull interfaces to various monitoring tools (e.g. Nagios, Zabbix) and to OpenStack projects -> fast failure notification Mapping between physical and logical failures Expose more faults and changes to resources (deduced alarms and states) Provide Root Cause Analysis indicators to the application manager Can be configured differently for different systems
17
OpenStack Congress
18
What is Congress? Governance as a Service Policy
Define and enforce policy for Cloud Services Policy No single definition Law/Regulations Business Rules Security Requirements Application Requirements Any Service, any Policy
19
Congress Architecture
Policy Data Policy Policy Enforcement Congress API Policy Engine Nova DataSourceDriver Neutron DataSourceDriver Keystone DataSourceDriver Security System DataSourceDriver Nova Neutron Keystone Security System
20
Requirements and gaps for Congress as Inspector
Congress Features Gaps Fast Failure Notification Periodical polling and policy enforcement Real-time policy enforcement Mapping of a physical failure to a logical failure Write a rule for mapping None Adaptability Change Policy rules
21
Congress PushType DataSource Driver
Policy Policy Another Service Policy Enforcement New data flow Congress Enables services outside Congress to push data, and improves reaction time for policy enforcement API PushType DataSourceDriver Policy Engine Nova DataSourceDriver Neutron DataSourceDriver Keystone DataSourceDriver Security System DataSourceDriver Nova Neutron Keystone Security System
22
Congress Doctor Driver
Data Policy 1. Policy Monitor Policy Enforcement New data flow Congress API 2. Doctor DataSourceDriver 4. Policy Engine 3. Nova DataSourceDriver Monitor notifies hardware failure event to Congress Doctor Driver receives failure event, insert it to event list of Doctor Data Policy Engine receives the failure event, then evaluate registered policy and enforce state correction Policy Engine instruct Nova Driver to perform host service force down and reset state of VM(s) Neutron DataSourceDriver Keystone DataSourceDriver Security System DataSourceDriver Nova Neutron Keystone Security System
23
Congress Doctor Driver (Detail)
Driver Schema(HW failure example) | table | columns | | events | {'name': 'id', 'description': 'None'}, | | | {'name': 'time', 'description': 'None'}, | | | {'name': 'type', 'description': 'None'}, | | | {'name': 'hostname', 'description': 'None'}, | | | {'name': 'status', 'description': 'None'}, | | | {'name': 'monitor', 'description': 'None'}, | | | {'name': 'monitor_event_id', 'description': 'None'} | Event List of Doctor Data | id | time | type | hostname | status | monitor | monitor_event_id | | ab | T07:39: | host.nic1.down | demo-compute0 | down | demo_monitor | |
24
Doctor Wiki: https://wiki.opnfv.org/display/doctor
Visit Doctor Booth!
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.