The cornerstone of automated operations: CMDB

the cornerstone of automated operations: CMDBOriginal 2016-03-29 Yu Junwei Broad-pass software agile operation and maintenance

Today, for everyone to highlight March 25-26th Shenzhen Global operation and Maintenance Conference, automation operation and maintenance of excellent cloud product director Yu Junwei classmate speech of PPT dry (at the end of the article can be downloaded). "The cornerstone of automation operations: CMDB" keynote speech, the scene response is very good, is also a lot of participants in the Shenzhen Global operation and Maintenance conference hope to exchange and learn one of the themes.

Yu Junwei classmate, from Guang Tong Software, is responsible for the excellent cloud operation and maintenance product line research and development, is an IT zoo in the more than 10 years of Operation Engineering Lion + program ape.

The global operation and maintenance conference These two days, many lecturers mentioned the CMDB, the importance of the CMDB, so for the automated operation of the CMDB, make a special share.

What is a CMDB.

While operations and prosperity are flourishing, fragmentation issues arise, and everyone wants to integrate the operations platform, but it often backfired.

A CMDB is like the core of a person's brain, a library of information that stores information that coordinates the body's complete variety of complex movements.

The CMDB in my heart

. Fragment Consolidation

Fragmentation scenario for operations tools is the core of data that revitalizes the entire operation and maintenance management

. Meta database

Provides the basic metadata for operational activities and is the only trusted operational configuration data service

. Scenario Driven

Data-driven for operations linkage, coordinating tools to complete a wide range of automation scenarios

Automatic expansion + automatic monitoring

How the CMDB is built.

Pain point phenomenon and countermeasure I model is not well built

Problems that exist:

. Model granularity out of control

If the granularity is too small, even the network cable, memory bar becomes a configuration item, the last 70% data stored in the CMDB has no effect, but did a lot of useless.

. Lack of industry practice reference

Many times in China are based on the BMC, HP and other models to build a model library, but in fact, the idea of foreigners and people are very different, often will make too complex model system.

. Model Adjustment too cumbersome

With a relational database, each type of attribute in the model is a column, and the final adjustment is always to use development, the completion of an adjustment takes 2 days, and this adjustment in the data replenishment phase, often to be carried out, time consuming.

How we do it – management

. Target-driven

Continue to iterate in a way that only implements the minimum set of models required by the current target. It is not recommended to use the traditional software to develop a waterfall model to build a model, but to use a continuous iterative approach, each time set a small target, according to this goal to build just meet the requirements of the Model library.

. Industry reference

Find and learn from industry best practices. Look for best practices in the industry, learn their models, and especially learn their evolutionary path, and don't eat a fat bite.

How we do it – technology

The first step, data type labeling, support multi-identity

Traditional CMDB systems, often using scientific classification of ideas, according to the boundaries, doors, outlines, and other tree structure to strict division, but this has brought great challenges to modeling, because there must be some data sibuxiang. For instance, whether the virtual machine is zoned to the traditional computing equipment resources or the virtual resources. So we propose to classify them in a way that is labeled with data types. For example, a virtual machine, I can simultaneously hit the computing device and virtual resources such as two tags.

The second step is to use relationships to establish relationships and distinguish between attributes.

A relationship that uses a weakly typed constraint, not a property. Because properties tend to be modeled ahead of time, many configuration items are actually created so that it is not clear which configuration items it might be associated with, so the use of relationships can be more lightweight.

Third step, easy to adjust the model, support dynamic properties

In the technical design of the CMDB system, it is important to pay attention to the use of fast-tuning storage models, such as using a scheme-friendly database, or a database such as PostgreSQL, which supports JSON extension fields, to implement dynamic properties.

Pain point phenomenon and Countermeasures II inaccurate data

Problems that exist:

. Manual data entry, low accuracy

. No timely maintenance, data expiration

. There are many sources of data, conflicts

How we do it – management

. Identify status

Determine the CMDB as the only data source, if the upper and lower data stream is inaccurate, should be corrected from the CMDB

. Delineation of authority

Self-setting principles, such as who provides, who maintains

. Periodic review

The system needs to be determined that the team can periodically audit data in the CMDB, look for erroneous data and improve the problem. As with some warehousing management, it is necessary to periodically check the book and physical inventory, and the CMDB also needs to periodically review the actual compliance of the data with the production environment.

How we do it – technology

. Support Collaboration

Configure the change hotspot and subscribe to the configuration item changes that I care about. Everyone can view the data footprints of others, the configuration items also allow to make a hot map by the number of changes or the number of times used, and finally should be allowed to subscribe to the configuration items I care about, so that when the configuration item changes, the relevant owner can receive timely notification.

. Record history

Allows you to query the history of your data at any time, and to backtrack to baselines. Each time the data is put into storage, it is possible to record the change history of the data so that it can be compared to the version changes at any time, as well as to retrace the baseline when correcting.

. Support Reconciliation

The use of policies, rules to achieve the harmonization of multiple data sources. Too many data sources can lead to data conflicts. In the event of data conflict, can show the conflict of different data sources, and support artificial reconciliation, while the CMDB system should also learn these artificial harmonic basis, can form an automatic reconciliation.

. Dependent tools

In the data collection and the supplement, uses the monitoring and the automation tool mainly, they can reduce the massive entry work, and avoids the human error.

Pain point phenomenon and countermeasure III data is not useful

Problems that exist:

. I don't know what the usage scenarios are.

There are often situations in which the CMDB is used for the CMDB, resulting in the last CMDB just as the resource ledger, and the most commonly used function becomes only Excel import and export. In fact, what we need to build is a service-oriented CMDB.

. Poor system openness

The CMDB is poor in openness, and often only provides a read-write API to use the CMDB as an ordinary database.

How we do it – management

Proactively look for scenarios, consume data, and generate value for your data.

Impact Analysis: Use message disk, do configuration change walkthrough, do failure Walkthrough.

Automatic monitoring: When you add some configuration items, you can notify the monitoring system and automatically generate a monitoring strategy.

Automatic troubleshooting: Can be automatically disabled when a fault is detected.

Capacity management: Expand the admission threshold for application records in the configuration library to automatically scale the expansion.

Transport: Data in the CMDB, in the current mobile terminal scenario, there is a particularly good consumption scene, is to do two-dimensional code, RFID, and mobile phone combination, can be in the room inspection and troubleshooting in the production of great convenience.

How we do it – technology

Relational derivation: Provides the ability to refine other configuration items by relationship from one configuration item.

Full-Text search: Easily use keywords to search for compliant configuration items.

Change Notification: Configuration item changes not only provide notification to people, but also use MQ to provide notification of operations tools to trigger some automation scenarios.

Transaction control: Allows the sandbox to be built through the API, and the configuration items in the entire sandbox are committed together and rolled back together, which is particularly useful for applications that go live.

Version comparison: Allows querying of historical data and changes of a configuration item.

Web integration: In addition to APIs, it should also provide interface integration between applications and should provide interface integration between applications as well as interface integration between applications.

CMDB Success Factors

The CMDB that can be consumed is a good CMDB.

Model: Defines the minimum available CMDB model structure and rules

Data: Correctly maintains the data of the CMDB and its relationship

API: Provides an open and friendly API service

Scenario: Using data from the CMDB to play a variety of operational scenarios

CMDB = model + data + API + scenario