MuleSoft Certified Integration Architect Exam (Latest 2024/ 2025 Update) Questions and Verified Answers |100% Correct| Grade A

MuleSoft Certified Integration Architect
Exam (Latest 2024/ 2025 Update) Questions
and Verified Answers |100% Correct| Grade
A
Q: What are typical types of documentation in an integration solution architecture?
Answer:

Required use cases
Views of systems and sub-systems
Views of data and interfaces
Functional requirements and non-functional requirements and SLAs
Views of interactions and design decisions
Key decisions, requirements and tradeoffs
Q: What is the 4+1 methodology?
Answer:
Is one common approach how to illustrate views for software and systems

Logical view: Architecture diagram, relationships between systems and stakeholders
Process view: API-led diagram, Information about the movement of information, Documents
success and failure paths, Sequence and activity diagrams
Development view: Data mapping, testing strategy, Example flows in Mulesoft
Physical view: Deployment diagrams, What artefacts are running on each node, how the
different pieces are connected, high level topology: databases, runtimes, workers, LBs, firewalls,
VPCs
+1 = The plus 1 illustrates the scenarios from an end user perspective –> User Stories

Design phase: Process view, development view, physical view

Q: What documentation is also required beside 4+1 view?
Answer:
Maintenance, operations and security documentation
Q: What are Mule applications under the hood?
Answer:
Java-based applications based on Java Spring configured by Mule application XML files
Q: Which 4+1 view is created to direct deployment and maintenance phases?
Answer:
Physical view:

Shows the process and the nodes that are running
View on CI/CD
Q: What is a Mule application?
Answer:
It runs in a Java virtual machine (JVM)
It is triggered by internal or external events or processes and routes to other components or
endpoints
Q: How does a typical Mule flow looks like?
Answer:

Q: What are the main elements for a Mule event?
Answer:
Mule events can be passed or copied between event processors in the flow

Message: The main data object carried between event processors
Attributes: Metadata contained in the message header
Payload: Core information of the message
Variables: Stores current event state for use by later event processors
Error message: Generates by the flow
Q: What are the three types of flows in Mule 4?
Answer:

Regular

Starts with a message source
Can have its own error handler

Private flow

No message source
Can only be triggered from within the Mule application

Sub flows

No message source like a private flow
No Error handling (Errors bubble up into the parent flow)
Q: What happens if you route a message with a flow reference to a sub flow?
Answer:
The flow message is processed synchronously
Q: How can you trigger flows asynchronously from a parent flow?
Answer:

With the Async scope:
It will copy the message and process it asynchronously in the parent and sub-flow
For example if you want to send e massage to Splunk you can use the async scope
Q: With which component can you trigger flows beside an event source?
Answer:
Scheduler:

Fixed frequency: Poll every 1000 mili sec
Cron: You can specify an event to occur just once at a certain specific date or at some
frequency
When you enable CRON you can see the application on cloud hub an trigger it in the run time
manager
runtime manager -> schedules
You can also trigger flows directly from connectors with internal schedulers: Database, File,
FTP, SFTP on Table Row and on New or Update File
Q: Which connectors have internal scheduling options?
Answer:
Database: on Table Row
File, FTP, SFTP: On New or Update File
Q: What are the advantages of using connectors?
Answer:
Facilitate integration of third-party systems such as Salesforce, Google, Facebook, Workday
No need to study and code to underlying protocol or security by the third party system
Powered by https://learnexams.com/search/study?query=

What characterises integration solutions across an enterprise?

Various systems and applications to connect
Various stakeholders
Conflicting goals
Stakeholders have different assumptions, understandings and language
Systems have different requirements, reliability, availability and performance

What is the most important objective of an integration solution?
To address the in-scope requirements of stakeholders related to the scenarios and use cases

Identify stakeholders involved in an integration project?

Project sponsor: drives the project
Architects: Responsible for implementation
System integrators and external stakeholders: responsible for external systems
Auditors: Verify compliance, policies, integrity
Users: Enduser of the system

Identify stakeholders related to Anypoint Platform?

Users
Administrators
Developers and Architects
System integrators and external stakeholders
Analysts and Managers: responsible for managing the scope, value, costs etc.

What are the main areas which an architect has to deal with?

Non-technical business vs. technical stakeholders
End users
Development vs. deployment vs. runtime operations stakeholders
Implementers vs. managers. vs. executives

What is Mulesofts point of view on integration solution architectures vs. enterprise architectures?

Integration solutions architecture: Documents individual integration initiatives for example message-based integration, batch processing, ETL, form of integration between systems
Enterprise architecture: Application network of an organisation. API-Led connectivity, layering, reusing, combining API-Led

What is Mulesoft’s OBD = Outcome based delivery approach?

90% of the projects are not reaching their goals
After few months of production the customer can not use the platform because the Organisation is not enabled enough

Business outcomes = What do we deliver?
Technology delivery = How can we use Anypoint Platform?
Org enablement = How can we enable C4E and training?

How are user stories usually documented?

What are the two kinds of requirements which are discovered from user story perspective?

Functional requirements:

Triggering events
Acceptance criteria
Expected errors and error handling

Non-functional requirements:

May be invented by industry or other external authorities
Are constrains on the requirements

What are typical types of documentation in an integration solution architecture?

Required use cases
Views of systems and sub-systems
Views of data and interfaces
Functional requirements and non-functional requirements and SLAs
Views of interactions and design decisions
Key decisions, requirements and tradeoffs

What is the 4+1 methodology?

Is one common approach how to illustrate views for software and systems

Logical view: Architecture diagram, relationships between systems and stakeholders
Process view: API-led diagram, Information about the movement of information, Documents success and failure paths, Sequence and activity diagrams
Development view: Data mapping, testing strategy, Example flows in Mulesoft
Physical view: Deployment diagrams, What artefacts are running on each node, how the different pieces are connected, high level topology: databases, runtimes, workers, LBs, firewalls, VPCs

+1 = The plus 1 illustrates the scenarios from an end user perspective –> User Stories

Design phase: Process view, development view, physical view

What documentation is also required beside 4+1 view?
Maintenance, operations and security documentation

What are Mule applications under the hood?
Java-based applications based on Java Spring configured by Mule application XML files

Which 4+1 view is created to direct deployment and maintenance phases?
Physical view:

Shows the process and the nodes that are running
View on CI/CD

What is a Mule application?

It runs in a Java virtual machine (JVM)
It is triggered by internal or external events or processes and routes to other components or endpoints

How does a typical Mule flow looks like?

What are the main elements for a Mule event?
Mule events can be passed or copied between event processors in the flow

Message: The main data object carried between event processors
Attributes: Metadata contained in the message header
Payload: Core information of the message
Variables: Stores current event state for use by later event processors
Error message: Generates by the flow

What are the three types of flows in Mule 4?

Regular

Starts with a message source
Can have its own error handler

Private flow

No message source
Can only be triggered from within the Mule application

Sub flows

No message source like a private flow
No Error handling (Errors bubble up into the parent flow)

What happens if you route a message with a flow reference to a sub flow?
The flow message is processed synchronously

How can you trigger flows asynchronously from a parent flow?
With the Async scope:

It will copy the message and process it asynchronously in the parent and sub-flow

For example if you want to send e massage to Splunk you can use the async scope

With which component can you trigger flows beside an event source?
Scheduler:

Fixed frequency: Poll every 1000 mili sec
Cron: You can specify an event to occur just once at a certain specific date or at some frequency

When you enable CRON you can see the application on cloud hub an trigger it in the run time manager

runtime manager -> schedules
You can also trigger flows directly from connectors with internal schedulers: Database, File, FTP, SFTP on Table Row and on New or Update File

Which connectors have internal scheduling options?

Database: on Table Row
File, FTP, SFTP: On New or Update File

What are the advantages of using connectors?

Facilitate integration of third-party systems such as Salesforce, Google, Facebook, Workday
No need to study and code to underlying protocol or security by the third party system
Speed up development and deployment
Make mule applications easier to maintain

How can you share global elements across Mule applications?
Mule Domain project:

Mule domains are only supported in customer-hosted runtime planes
One or more application can be associated with the Mule domain
A Mule domain can be deployed to Mule runtimes along with Mule applications

What is important about Mule Domains when you want to use them?
They are only supported in customer-hosted runtime planes

What changed in Mule 4 compared to Mule 3?

Messages are now called events and the structure has changed
Inbound and outbound properties are called attributes
Each Mule 4 connector explicitly sets its own outbound properties
Session variables are removed and instead each connector passes data over transports as Payload, Headers attachments
Variables are now stored and passed around with the Mule event

How did flow change on Mule 4?
We don’t use processing strategies anymore in Mule 4. Everything is auto tuned

What has changed between Mule 4 and Mule 3 looking at the connectors?
In Mule 4 the connectors are not bounded anymore to the runtime

Each connector and many other components are separated out into independent extension libraries called modules

In Mule 3 you had to wait for the new runtime if a new connector was released

How are modules configured in a Mule application?

Each module is configured as a Maven dependency in the pom.xml file

What is supported by the HTTP connector?

Inbound requests over HTTP and HTTPS to trigger a flow
Send an HTTP or HTTPS request in the middle of the flow
HTTP 1.0 and 1.1

It does NOT support HTTP 2.0

How are Grizzly libraries used for HTTP connectors?

Grizzly is a standard open source library supporting HTTP/S protocols
Uses selector threads pool
Listeners have a shared selector pool for each Mule app
Requesters have a dedicated selector pool for each Mule app

What can you do if a connector is not available?
You can build your own custom connector with Mule 4 SDK

What are the 4 Anypoint Connector types?

Community: Mulesoft or Community written, no license
Mulesoft Certified: Developed by Mulesoft partners and developer community and reviews by Mulesoft
Select: Mulesoft maintains select connectors, To use them you must have an Anypoint Platform subscription
Premium: Premium Connectors like SAP, you need an active CloudHub premium subscription

How does the connector reconnection strategy works?

By default a failed connectivity test is just logged and the application starts anyways or continues to run without trying to reconnect
A reconnection strategy can be configured to try repeatedly to connect
A failsDeploy attribute can be set to true to throw an exception if the reconnection attempts fail which prevents the Mule application from starting

What do you need to consider when writing DataWeave?

Use functions like filter in order to improve the performance and do not iterate with loops in order to filter
Always use Dataweave type: application/dw in development phase. its easier to debug because its better to read the errors
At the end you can change the type to application/json or whatever you need

What are types of Mule routers?

Choice
Scatter-Gather
First Successful
Round Robin
Error handling

What are characteristics of Round Robin?

Routes the incoming Mule event to one of its routes
A different route is selected each time in looping
Each invocation of the Round Robin router is synchronous

Round robin uses object store to remember the last execution

What happens if an error is not handled?
If an error is not handled the error message is logged and the flow processing stops

Where can an error handler be added in a Mule application?

Global Error Handler
A regular Flow
A try scope (sequence of one or more mule event)

An Error Handler can not be added to subflows

What is the difference between on Error Propagate and on Error Continue
Propagate:

all processors in the error handling scope are executed
At the end of the scope:
. The rest of the flow that threw the error is not executed
. The error is re-thrown up to the next level and handled there
If the error starts with an HTTP listener it returns Error 5XX response

Continue:

All processors in the error handling scope are executed
At the end of the scope:
. The rest of the flow that threw the error is not executed
. The result of the error handler scope is passed up to the next level as if the flow execution had completed successfully
If the flow starts with HTTP listener it returns 2XX response

What happens with an error handling within a transaction?

On error propagate:
The error is thrown up the call stack so the current open transaction is rolled back
Continue:
The error is handled so the current open transaction is committed (no roll back)

Example:
If you write a database on error propagate the previous transaction is rolled back

How does a try scope handle errors?
Acts like a private flow inside the parent flow:

Can have its own error handling
Unhandled errors bubble up to the parent flow

What is reactive programming?
A programming paradigm that combines the concurrency with event-based and asynchronous systems

Mule runtime is based on reactive programming

Example: You get a lot of events and you perform action as soon as you want to perform them
(Pizza example)

What is back-pressure and how is it handled by Mule?
When Salesforce is slowly, the connector will slow down as well

What processing model is used by the Mule 4 runtime?
A non-blocking and reactive processing model:

Non blocking is a central theme in reactive principles
Non blocking is the norm in Mule 4

What does non-blocking mean?

When you make an outbound request the thread will be released back to the grizzly-thread-pool until receiving an answer

What happens if the scheduler interval is faster then the processing of the flow? What do you need to do in this case?
Mule runtime throws an error and loses the message.

You need to set max. concurrency in the flow to: 1

How many global executors do exists to run the tasks in the Mule 4 runtime?
3

What are the three execution types that can be set in an Anypoint connectors source code?

IO_INTENSIVE (Blocking)

Transaction scope
all blocking IO intensive connectors
All technique that blocks the current thread

CPU_INTENSIVE

Intensive processesing
Dataweave

CPU_LITE

Processing that neither blocks or is CPU intensive
Routing, filtering, message passing

What is RESTful?

Addressable resource:
A resource is any information or concept that can be named
Stateless:
The server does not store application state
Connectedness:
The server guides the client to change state –> Done by Hypermedia
Uniform interface:
Resources share interface characteristics
Idempotency:
The same HTTP request should always create the same server state

Which HTTP methods are idempotent?

Read
Updated
Delete

Idempotent: It doesn’t change the results no matter how ofter you klick it

What is the difference between SOAP and REST?

SOAP focuses on exposing a piece of application logic (not data) as service
SOAP is based on the XML and W3C standard
SOAP and REST support SSL
SOAP focuses on exposing and accessing named operations, while REST focuses on accessing resources via HTTP methods

What is special in Mule 4 talking about streaming? What is streaming?
The Mule 4 automatically streams large data payloads without any special configuration

Steams are data structures which are processed continuously as it arrives

Data will hold in memory before computing

Which connectors support streaming?

File
FTP
DB
HTTP
Sockets
Salesforce

What are the 3 streaming options for connectors in Mule 4?

File stored repeatable streams:

By default stores 500 objects in its in-memory buffer (number can be configured)
Excess objects are serialised using a Kyro serializer

In-memory repeatable streams

Max buffer size 500 objects
If streams exceed the max bugger size then the application fails

Non repeatable streams:

The input stream is only read once
No extra memory or performance overhead compared with repeatable streams

What is an asynchronous processing model?

The main thread does not wait for its workers to complete execution
Uses branch processing
Failure in one of the branches does not impact the main flow
Responses from branch processing are not available to the main flow

How can you achieve with a JMS connector synchronous and asynchronous communication?
Synchronous:

Publish sends a message to a queue
never gets a response back to the message provider

Asynchronous:

Publish consume –> Blocks the flow until a response is returned from the JMS provider or the timeout expires

What is the difference between JMS and VM queues?

Unlike JMS, VM queues do not use any intermediate message broker
Creates and communicates with virtual machine using publish/subscribe module
Supports sync and async
Supports intra-app and inter-app (in Mule domain) communication

How do you persist data with a VM queue?

On customer-hosted Mule runtime: by serializing and storing the content on the disk
On cluster of customer-hosted Mule runtime: persistent queues are backed by the Hazelcast distributed data grid
On CloudHub: Are stored in a CloudHub service

What is the difference between VM queue and other queues?

With VM queues reliability and quality is limited
VM queues can be used for specific use cases:

To distribute messages across a cluster of Mule runtimes
For high-performance async communication
To distribute work within the same Mule application across multiple workers
When investment in a JMS broker is not supported (VM is for free)

Describe the batch job processing model?
(only available with Mule Enterprise runtime)

From the input payload a fixed block of records is send to the batch job
All batch records are first queued
Records are taken from the top of the queue one at a time and sent to the first batch step
Several threads may process multiple records in parallel
All batch records will be queues and processed again for batch step 2

Important: It can be that some records are still processed in step 1 and some are already taken to step 2 if threads are available in the pool

What are batch step filters used for?
You can define an “accept Expression” if a record did not fail during the previous batch step

You can configure another filter which handles previously failed records

What are characteristics of BatchJobResult?

Contains a summary report about the records processed for the particular batch job
Does not include any of the actual processed records or data
The payload after a batch job is the original payload before entering the batch job

Which API data types are supported by Mulesoft?

JSON and XML
REST API RAML data types
OpenAPI OAS data types

How can you convert data types from RAML to OAS?
using online tools: apimatic.io

or its done automatically by Mulesoft. Full conversion will be available in the new release

What data types are supported by DataWeave?

What is important to know about Java and DataWeave?

Java is generally NOT recommended for data transformations
Also other languages: Ruby, JPython, Nashoron etc is not recommenced tu use

However

DataWeave can call out to static methods in Java
Java classes can be defined in Spring context
Mulesoft recommends to encapsulate custom Java transformations in classes
Mulesoft promotes separation of concern hence removed the expression component

How do you enrich messages with external payload?
In Mule 4 components have Target and targetValue.

A target saves you the step of having to store the previous event payload using a variable event processor
The payload from the previous event is then left unchanged
The target can also operate on event processor result

What is a common data model and why should it be used?

To represent data in new applications that need to communicate with other internal applications
When the process of defining a common data model does not delay implementation indefinitely

In Mulesoft you would have:

1 transformation on experience API to the CDM
1 transformation on system API from CDM to external system

What are the scopes a CDM can take place?

Just the current project
To the business unit
To the entire enterprise

What are ways to implement data validation patterns in Mule applications?

Choice routers
DataWeave Code
Validation Modules
Catch and Handle Errors

What modules provide validation operations?

Validation Module
JSON validate schema
XML validate schema
APIKit, JSON validation, throws SOAP faults
Java validate type
HTTP request operation, validate response, identify success vs. failure status code

How can you combine validations?
With the All scope

How can you use the validation module without throwing an error?

Validation operations can be used directly inside DataWeave code
In this case the operation just returns true or false and does not throw an errors

You can use this to write choice router conditions

How can you configure sequential execution in a Mule flow?

set max concurrency of 1

This ensures that only one flow instance is processed at any point

For each scope

splits a payload into elements and processed them one by one through the components that you place in the scope

For loops

Whar are collections in for each?
Collection = What are the iterable colletions

By default for each tries to split the payload into iterable types
If the payload is a Java collection it can be done without any configuration
If the payload is non-Java you need to specify the iterable collection to use

Why is parallel processing sometimes required?
Problem:

In a single flow processors are executed in sequence and each processor execution is dependent on the execution of preceding processor
The processor that communicates with an external system might block processing of the current Mule event

Which processors provide parallel processing?

Scatter-Gather
Async scope
VM
JMS

What happens if a scatter-gather throws an error?

if type: MULE:COMPOSITE_ROUTING
The error contains the result of every route
Event processing does not continue with the next event
Instead the flows error handlers process the error as they would any other error type

What is important for async scopes and what happens with errors in an async scope?

Response from the async scope is NOT accessible nor is it returned to the main flow
Events are processed inside an Async scope concurrently with the event from the main flow

Errors inside the async scope do not impact the main flow

You can use try scope to customize error handling

Which types of testing are relevant for Mule applications?

Unit testing: unit specific, functional
Integration testing: all interactions
Performance testing: Scalability, Reliability, Usage, Workload

What is Before Test in Munit?

You can set the Mule event passed to the execution scope

How can you configure that only certain tests are executed?
You can add tags to the Munit tests

What are populate tools for blackbox integration testing?

SOAPUI
Restlet Client
REST-assured

What tools can be used for performance testing?

JMeter
BlazeMeter

What are the different deployment models that Mulesoft provides?

Mulesoft-hosted runtime plane –> cloud based
Customer-hosted runtime plane –> provisioned and managed by the customer

For non mulesoft hosted environments like AWS, Azure, PCF
Also on-prem infrastructure, on bare-metal, virtual machines or containers like docker

What are the features of a Mulesoft-hosted runtime plane?

A new virtual machine is automatically provisioned for the new Mule runtime –> called Cloudhub worker
A Mule application can be scaled vertically and horizontally
Mulesoft provides a load balancing service
Mulesoft implements a distributed object store service

What is horizontal and vertical scaling?
Horizontal: The Mule application can be automatically deployed to multiple Cloudhub workers

Vertical: A Mule application can have its worker resized to a larger or smaller vCore size

What is the underlying infrastructure of Cloudhub?
EC2 AWS instance with a Mule runtime with a shared load balancer

–> CloudHub is on AWS

What are Mulesoft iPaas options?

Cloudhub
Runtime Fabric
PCF

What is Runtime Fabric?
It is an additional software provided by Mulesoft to scale out customer-hosted Mule runtimes

Mule applications deploy to isolated Mule runtime which run in a docker container
The containers can be deployed to any cloud service
Provides additional services like zero-downtime redeployment and load balancing

What are the differences between Anypoint control plane on Mulesoft-hosted or Private Cloud Edition?

What are advantages of a Mulesoft-hosted runtime plane?

Each mule runtime is automatically installed in a separate CloudHub worker
No contention for host resources
Automatic upgrades/patches
Globally available
Mulesoft provides security and SLA
Managed by Mulesoft hosted anypoint platform control plane

What could be a reason for the private cloud edition?
For customers with strict regulatory or compliance requirements

What is the difference between Private Cloud Edition and Pivotal Cloud Foundry?

PCE vs. PCF
Same as Anypoint platform except the runtime plane is deployed on PCF instance

–> Like CloudHub, one Mule application per Mule runtime

How does runtime fabric works?

Runtime fabric orchestrates and automates the deployment of Mule runtimes into containers in any cloud or on premises
Deploy consistently across any cloud or data center
Run multiple Mule runtime versions in the same runtime fabric
Isolate apps, scale horizontally, redeploy, zero downtime
Connect to the control plane hosted by Mulesoft
No need to dockerize Mule apps

What are the main differences between CloudHub (CH), Runtime Fabric and Customer-hosted?
Inbound HTTP load balancing: CH, Fabric
No inbound HTTP load balancing: on-prem

VM queues messages are load balanced for deployments to multiple CH workers: CH

VM queues messages are load balanced for deployments to a cluster of Mule runtimes: Fabric, on-prem

Rescale Mule runtimes with zero downtime: CH, Fabric

no rescaling: customer-hosted

What could be factors for deciding to a on-premis option?

Regulatory
Time-to market
IT operations efforts
Accessing on-prem data sources
Flexibility of deployment across cloud providers
Isolation between Mule apps
Control over Mule runtime tuning
Scalability of runtime plane
Roll-out new releases
Redeployment with zero downtime

Please provide for each use case appropriated deployment options:

Regulatory
Time-to market
IT operations efforts
Accessing on-prem data sources
Flexibility of deployment across cloud providers
Isolation between Mule apps
Control over Mule runtime tuning
Scalability of runtime plane
Roll-out new releases
Redeployment with zero downtime
Regulatory: PCE, PCF
Time-to market: CH
IT operations efforts: CH or Fabric/PCF over PCE
Accessing on-prem data sources with high throughput: Fabric, PCE, PCF
Flexibility of deployment across cloud providers: Fabric, PCE
Isolation between Mule apps: favours bare metal over VM or containers
Control over Mule runtime tuning: Hybrid –> Fabric and PCE
Scalability of runtime plane: CH, favours automatically provisioning then manual
Roll-out new releases: Weekly releases for CH and quarterly for PCE
Redeployment with zero downtime: CH or Fabric

What are use cased for saving state in a Mule application?

Between flow executions:

To remember state from previous outbound HTTP request
To filter out already processed messaged (watermark)
To cache unchanged data or responded to speed up response time

Between iterations when processing a collection of elements:

Including batch processing
Including iterations driven by a scheduler or cron job

What is the difference between persistent and non-persistent data store?
Persistent: Data survives when redeploying or when server crashes or CH worker is stopped

Non-persistent: Data can be reused while the application is still running but data is lost after application restarts or server crashes

What are the options to store states in Mulesoft and what is the fastest?

Fastest: non-persistant
persistant: latency to write data to the local file
state replicate to a distributed memory data grid: added latency to replicate across network
back a mule runtime cluster with persistent database storage: latency to replicate data across the network

What is an object store and what is it designed for?

Mulesoft object store is a key-value store implemented in Java and accsesed via Object Store connector

It is designed for:

Synchronization information like watermarks
Temporal information like access tokens
User information

What does Mulesoft understand by persistence?

persistence: storage that is copied to a disk or some externals storage or replicated across several nodes
non-persistent: usually refers to data that is only stored in the JVM memory

Each object store can be configured as persistent or transient (non-persisten)

What does Mulesoft mean by durability and reliability for Object Store?
One vs. multiple CloudHub workers
One vs. multiple clustered customer-hosted Mule runtimes

All these factors affect the overall performance and SLAs of an object store

How does a Mule application implements an object store?
Persistent: OSv2 = Anypoint Object Store service

cloud native implementation
Data is shared between all CH workers

Non-persistent: Does NOT use OSv2 service

object store is implemented locally in each CH worker
Data is isolated within each CH worker

What are the two options to configure performance of a Mule runtime cluster?
You can set:

reliable or
performance

by default its set to reliable

What are the limitations of Anypoint Object Store?

Not a universal data storage solution like S3 bucket
Does not support transactions
Does not replace an actual DB and not suitable for searches or queries
Data is automatically removed after 30 days for OSv2
Max storage per message is 10 MB

Where is the data saved depending on the runtime plane type?
CH:

Data is saved when the Mule application is deployed to CH
To shared distributed memory when the Mule app is deployed to a cluster of runtimes

Customer-hosted:

Can use OSv2 REST API to store and retrieve data

How can you block access to an object store from other Mule application components?
A private object store can be configured by a particular component to securely hide its object store data from another component in the Mule application

What are the object store types that Anypoint provides?

Global object store:

can share state between components in a mule application
For OSv2 between workers of a multi-worker CH deployment

Private object store:

When sharing data between components in the flow is deemed to be security risk

What are the characteristics of VM queues used to store the state of an application?

Queues can be transient or persistent
Transient queues are faster than persistent queues but less reliable

On customer-hosted runtime plane:

standalone mule runtimes persistent queues are stored to the local disk
cluster of mule runtimes, persistent queues are backed by the clusters distributed data grid

What are the advantages of VM persistent queues instead of another 3party messaging solution?

Use persistent queues when the use case does not require
More durable and reliable then JMS or DB
To share between multiple applications or with non-mule applications
persistent queues are not available

What is the difference between JMS and VM?

VM no publish subscribe
VM not external
JMS expensive

Difference for persistent queues:

You can not see the message in a VM queue

Why should you not use file-based persistence for Mule application on CloudHub?

Mule applications have only limited and ephemeral file system access
EC2 disc storage is removed when CH worker from the EC2 instance is removed
The file connector can only access specific folders such as /temp or /opt/storage

When should you use an external store to store and manage Mule application state?
Pros:

May provide more exacting performance, reliability, durability
May be preferred option to achieve certain high availability or failover goals
Use when backup and replication is required for cache objects

Cons:

Additional cost, management and staffing requirements
Added layers and complexity

What is a watermark?
A watermark is a form of state that is stored during a recurring processing cycle such as to process a collection of records

Mostly used to avoid dublicates

What types of watermarks can be used?
Automatic:

saving, retrieving and comparing is automatically handled through an object store
Only for few connectors available:
on new or Update File or On Table Row

Manual:

Handle saving, retrieving and comparing the watermark
More flexible in that you specify exactly what records you want to retrieve

How can you send logs to an external logger?
With an additional appender in log4j

Supports synchronous and asynchronous logging
By default logging in Mule is done asynchronously

What is the difference between System and Application logs?
System log:

specific to the mule runtime
configured by log4j and inaccessible to customers

Application log:

specific to the mule application
configured by log4j typically packaged in mule application

What is the difference between synchronous and asynchronous logging and when is it used?
Synchronous:

Thread which is processing the message is interrupted to wait
Used when log is used as an audit trail or ERROR/CRITICAL message

Asynchronous:

Logging occurs in a separate thread so the actual message is not delayed
Log may be lost in case of system crash

How can you trace messages in Mulesoft?
The correlation id is automatically logged on cloudhub and customer-hosted mule runtimes

the logger component automatically logs the correlationId for DEBUG level messages

What are the retention limitations in CloudHub?

Up to 100MB per Mule application and per worker
At most 30 days

How can you enable if you want to send CloudHub logs to external logging systems?

With a custom log appender
You need to request via the Mulesoft support portal

What is another option instead of an additional appender in log4j to send data to an external system?

Custom aggregator application
Lambda is taking over the CloudHub APIs the logs and send it to another system

Where can you configure alerts and how can they be triggered?
Runtime manager

Can be triggered by Mule application or server conditions
Alerts can send notifications to email addresses
Custom notifications can be generated by a Mule application

What are business events and what are they used for?

Mule applications can collect business events information as each Mule event transitions through a flow
Business events are designed to collect KPIs and store them in an Mulesoft provided online service
The process may degrade network or Mule application performance

How can you enable Business Events?
Default business events can be enabled or disabled per flow

What is a Custom Business Event?

A Custom Business Event component can generate additional business events at a particular place in a flow
You can define a KPI with its related Value
KPIs can be stored outside the Mule application

What types of environments does Anypoint provide for deployment?

Production quality environment

Deploy apps and APIs to publicity
-. When you create a new account it contains one prod environment

Sandbox quality environment

Env for development and testing

Design quality environment

Enables you to test and run mule applications at design time

The underlying infrastructure is different depending on the type you select

Each VPC is assigned to a business group

How does a best practice CI/CD process looks like for Mule?

What are the options to run applications beside manually on CloudHub?

Anypoint CLI
Platform APIs

With which tools can you run static code analysis?

Githook
Sonarcube

What are the characteristics of a business group?

Business groups provide complete isolation of resources
vCores are assigned to a specific business group: Makes those vCores only available to the business group and unavailable to the parent organization
Each business group has its own environment
Each business group has a separate cliend id and client secret
Deleting a business group is NOT recoverable as all resources get deleted

What are the Maven lifecycle phases?

compile: compile mule source code of the project
test: run munit tests associated to the project
package: packages the project into a mule deployable jar
install: sends the distributable to a local repository ($MULE_HOME)
deploy: sends the distributable to a remote repository

What is a transaction?
Is a grouping of operations what are guaranteed to all complete or none at all

What is the two phase commit protocol?

First you check if the system is ready to commit
When its ready to commit, you commit to the DB

What is the XA interface?
If you have global transactions involving multiple separate resources such as database you need to manage it by a transaction manager.

XA-Interface specifies communication between a transaction manager and a resource manager

What are the alternatives if an API doesn’t support the 2phase commit?

Large distributed transactions systems: Google Spanner
Eventual consistency
Sagas
CAP theorem: two of them can be achieved: consistency, availability, partition tolerance

What are the two SAGA pattern?

Event

Each API creates an event and send it to the next one
No API is responsible for all

Command/Orchestration

If you have 3 APIs, one generates an event then send it to another payment API. This will receive and generate a success and send it to delivery
Every event is passed from API to API
In SAGA one API is responsible for everything. One Experience API is responsible for all others and if there is a failure the mein API will take the error

Which transaction types are supported by Mule?

Single-resource (local) TX

If you only have one database in the try scope

Global (XA) TX

If you have more then one database = distributed system

What are characteristics of Transactions?

Supported connectors must be configured to use transactions
All message processing done on a single thread
XA transactions need XA-capable transaction manager

Which connectors support transactions?

JMS
VM
Database

How do you begin and finish a transaction?
Begin a transaction:

Try scope
A transactional connector acting as an event source (Database, JMS, VM)

Committed automatically at the end of:

Flow
Try scope
On error continue scope

Rollback transaction:

After a failure occurs in a transaction scope, but only if the error is not handled in an On Error Continue Scope
On Error Propagate
By throwing an error in a flow or in a Try scope using: raise-error

Please check image and mention all transaction configurations

Transaction between databases –> Type = XA
Transaction action:

.Try scope: always begin
. Advanced properties: mysql join
. Advanced properties derby: join

What is the default XA transaction manager in Mule applications?
Bitronix

To use Bitronix declare it as a global configuration element in the Mule application
Each Mule runtime can have only one instance of a Bitronix transaction maanger which is shared by all Mule applications
For customer-hosted deplyoments define the XA transaction manager in a Mule domain

What are the different connector options for transactional behaviour?

How can reliability achieved in Mule applications?

until successful scope
reconnection strategies
redelivery policy
RETRY_EXHAUSTED exception scope
Transactions

How does until successful scope works?

It repeatedly triggers the scope components (including flow references) until they all succeed or until a maximum number of retries is exceeded
The scope provides option to control the max number of retries and the interval between retries

How do you configure a reconnection strategy?

How do you configure a redelivery policy?

How can you use RETRY_EXHAUSTED?
The connector raises an exception and published the message to the DLQ

How can you achieve reliability for non-transactional systems?

Zero message for non-transactional systems is achieved using reliability pattern
Splits processing between an acquisition flow and a processing flow
The flows do not call each other directly but use persisted queues

Describe the reliability pattern for non-transactional systems?
The reliability pattern consists of two flows:

Acquisition flow:

Receives incoming messages then dispatches them to the persisted queue
In case of failure until redelivery is exhausted the message is dispatched to a DLQ

The processing flow

Process messages from the processing queue then dispatches them to another persisted queue

What are HA and DR goals?
HA = High availability

How to keep the overall system operational when a system component fails

DR = Disaster recovery

How to restore a system to a previous acceptable state after a natural or man-made disaster

How can HA be achieved?

Horizontally scaling to multiple mule runtimes

Process on multiple concurrent physical machines/VMs/dockers

Load balancing and clustering:

Clustering uses an active-active model of node
Load distributes across the actives nodes

What is a cluster?
A cluster is a set of Mule runtime engines that acts as a unit.

A cluster is a virtual server composed of multiple nodes (Mule runtime engines)

The nodes in a cluster communicate and share information through a distributed shared memory grid. This means that the data is replicated across memory in different machines

On which runtime planes is clustering available?
It is not available on CloudHub. You can use Fabric in order do share or double scale your application and provide high availability

What are the benefits of clustering?
HA: Clustering guarantees HA. If a Mule runtime engine becomes unavailable due to failure or downtime, another node in the cluster can take the workload.

Performance: Clustering also improves performance since the nodes are processing parallel the messages.

Other benefits:

Automatic load balancing
You can set up alerts when a node goes down

What does unclustered load balancing for HA and performance means? (only available for customer-hosted runtime planes)

Multiple Mule runtimes are configured to run the same application
They DON’T share or synchronise data between Mule runtimes –> in clustering communication is shared over the shared memory grid
(you can use a shared database)
Messages must be distributed or load balanced with a third party product

How can you achieve HA on CloudHub?

Using multiple CloudHub workers –> more then 1
Workers do not share any memory
Workers can use an external system for state management
Each worker is created in a different availability zone in the same AWS region
Mule application can share and store data between workers with OSv2

What are the two ways how nodes (runtimes) can join a cluster in customer-hosted runtime plane?

Unicast: Cluster uses IP address for identifying server
Multicast: cluster group servers automatically detect each other

How do you share memory in a cluster?

Distributed Shared Memory Grid:

Hazelcast is used to create a distributed shared memory data grid
Data is automatically replicated and available between the clusters nodes
This allows data to survive if a node crashes or otherwise leaves the cluster

Components:

Components that use a cluster shared memory include: VM queues, Object Stores

Connectors:

Most connectors are not cluster-aware:
All connectors that use object store are cluster-aware
Cache scope, Idempotent message, Round robin

How does Round Robin stores the previous stage?
With an Object Store

What are the differences between Clustering and Load Balancing in a customer-hosted runtime plane?
Pro Clustering:

Shared, distributed memory
Ideal for HA scenarios
Build-in load balancing for VM queues
Built into Mule

Pro LB:

Easy to set up
No performance overhead due to data replication
Configurable load balancing

Cons Clustering:

Performance overhead due to data replication
Not supported by CloudHub
Required 3party product for HTTP load balancing

Cons LB:

Required 3party product
No data synchronisation
Manage idempotency

What do you choose if you want to have HA and performant processing. Clustering or would you rather go for LB?
In this case Load Balancing because due to the shared memory of clustering you need to replicate data and you have more overhead which is constraining performance

What is Anypoint MQ?
Publish Subscribe Service

What is the difference between VM queue standalone Mule runtime or clustering when it comes to load balancing?

VM queue standalone –> NOT load balanced

. Every standalone Mule runtime node will execute flow instances independently
. The processing happens on a single node
. No distributed processing

VM queue clustered –> load balanced

How can you identify performance bottlenecks?

Monitoring tools: anypoint dashboards, Jconsol, VisualVM, AppDynamics etc.

Why is application profiling performed?
Due to:

Memory issue

Detect memory leaks or excessive load situations
Java heap memory dump can help to analyze these types of issues

Application unresponsiveness

Detect blocked threads, long-running threads, waiting threads
Thread dumps can help analyze the issue

What are tools to monitor traffic and KPIs?

Anypoint Monitoring
New Relic
Splunk and ELK
Zipkin or Jaeger
Nagios

What are the scaling options you can apply on Mule runtime and Mule application?

Vertical scaling

Scale up –> More vCore size on CH
More Resources –> CPU, RAM on each machine

Horizontal scaling

Scale out
Process on multiple machines
Load balancing or/and clustering

How can you cluster for performance?

You can configure an Application or Runtime for high performance which is by default disabled
Mule Runtime:
Setting the storeprofile value in the mule-cluster.properties or wrapper.conf file: mule.cluster.storeprofile=performance
Mule Application:
You can also override per Mule Application in a configuration global element (screenshot)

What are the effects of setting the performance profile?

Disables distributed VM queues, using local VM queues instead to prevent distribution in the shared data grid
Implements node local in-memory object stores instead of shared memory grid architecture to avoid replication

Its not true that application always perform better in a cluster!

How does a Mule application deploys to multiple CloudHub workers?

CH automatically distributes multiple workers for the same Mule application across two or more zones
The HTTP load balancing distributes requests across all CH workers in an round-robin manner
A Mule app can scale out to max 8 workers or 16vCores

How can you enable autoscaling on Cloudhub and how is it configured?

Runtime manager provides an autoscaling feature
You have to ask Mulesoft to enable this feature
Allows you to scale to CPU or Memory usage threshold is exceeded
You can decide between:
. Increase vCore size – vertical
. Increase number of Mule runtimes – horizontal

Each autoscaling policy is triggered every 30min

What is the difference between a worker and vCore?
Applications on CloudHub run by one or more instances of Mule, called workers.

Each worker is a dedicated instance of Mule and can different memory capacity or processing power depending on the configuration.

You can scale workers vertically:

What are some configuration for designing a performant application? 10 topics

Synchronously

Better for performance because the throughput is less and small payloads

Asynchronously

For high throughput and large payload

Batch

Processes messages in batches (depending on the max batch size of external system – 200 messages)
High throughput

Scheduler

Scheduler runs periodically for new data
Useful for batch processing: use short interval to keep source and target in sync

Messaging queues

Decouple of data producers from consumers
Useful for real time data integration

Streaming

Data can be read multiple times or accessed randomly
Data can be sent to multiple places without need to cache data in memory first

Transformation

Avoid unnecessary conversions to Java
Let Mule handle streaming for you
Do not convert binary data types to Strings
DataWeave memory bugger of 1572864 bytes, if you exceed hard disk is used as buffer

Scatter-Gather

Parallel execution
Max concurrency sets concurrency for parallel execution of scope

Loggin

Async logging rather then sync

Network latency

Compress large payload (gzip)
Can use compression module
Caching responses

What is the default buffer size of DataWeave and what happens when you exceed it?
1.5 MB

It will use your system hard disk as buffer

What are some performance tuning considerations you need to think of?

HTTP or HTTPS connector

Always use HTTPS because the latency is not much worse

JMC connector

caching is on by default
Use sessionCacheSize to adjust size of JMS cached session
Disable JMS message persistent at JMS
Configure numberOfConsumers on JMS listener for high trough-put
Configure ACK mode to meet your SLA
Avoid durable subscriber and message filtering

Database connector

Caching on by default
Use bulk operations based on DB connector for batch processing
Enable streaming to start processing large results set
Max row and fetch size:
Example:
max rows = 1000
fetch size = 200 limits the response to max 1000 rows. in that it would require 5 separate network round trips

VM connector

Use for HA
Prefere flow reference within the same Mule app instead of VM endpoints

Batch jobs

Default block size 100
Max concurrency sets concurrency for batch jobs (default is twice the available cores in CPU)
Run comparative batch sizes to find optimum concurrency for the use cases
Use fixed site batch aggregator to do bulk operation for supported connectors

Streaming batch aggregator:

Receives all records in the job instance without running out of memory
SaaS provider often have restrictions on accepting streaming input

How can you tune performance on Mule customer-hosted plane?

JVM standard properties:

can be set in mule runtime weapper.config
Set initial and max heap size
Set NewRatio of the old and young generation heaps

You can use parallel GC = Garbage Collecting

What is heap memory and what are garbage collectors?
On customer-hosted plane:

Garbage collectors are used to move java objects from young generation heap to old generation heap

Java objects reside in heap memory
New objects are created in young generation heap memory
Garbage collectors move objects to old generation heap memory

What are the performance tuning parameters for flows and components?

Flow: max concurrency
Batch Job: Max concurrency, Batch block size
Batch aggregator: Streaming, Aggregator size
Scatter-Gather: Max concurrency
HTTP connector: use persistent connections
JMS connector: session cache size, cache consumer, cache producers, max redelivery
DB connector: Max pool size, min pool size, prepare statement cache size, transaction isolation
DB operation: Streaming, timeout, max rows, fetch size

Who is the owner of an organisation?
The person who creates it

What can you configure on high-level with the access control of Anypoint?

Organization: Created by the owner:
Business Groups:

Business groups has its client id and client secret
provides isolation of resources
vCores are assigned to business groups
vCores are only available for the BG and not to the parent organization
Each BG have its own environments
Deletion of a BG is not recoverable

Child Business Groups:
Roles
Environments
Users

Can you use external identity management server for Anypoint?
Yes, OpenID, SAML2.0

You can also apply oAuth2.0 policy to authenticate client applications

What secure communication protocols are supported by Mulesoft?

Symmetric and Asymmetric cryprography standards
Digital certifications

Protocols:

HTTPS
TLS
SFTP
FTPS
SMTP
IPSec

What is the difference in configuring Managing type in API Manager between:

Basic Endpoint
Endpoint with Proxy
Basic Endpoint:
Policies are downloaded to the application.
You apply the policies in the runtime itself
The policy is downloaded to /policies folder
API autodiscovery must be enabled to apply the policy

Endpoint with proxy: Proxy is applying the policy

If you are exposing non-mule applications
You have additional licenses

Can you apply additional policies on top of the Anypoint API policies?
Yes with Anypoint Security.

Its an additional product you need to buy
It offers you other types of policies that can be enforced at the edges of your network for multiple Mule Apps

How is edge security used by Mule applications?

Anypoint Security Edge is an enterprise solution for edge level security of APIs
It is a standalone product deployed outside Mule runtime
Its typically deployed in DMZ in a customer-hosted runtime plane

When is it applied?

To inbound request after they are sent from a calling API client before they arrive at the Mule application API endpoints (so before policies of API manager are applied)
To outbound responses back to the API client after leaving the Mule application API endpoint after API manager policies are applied

Are Anypoint Security policies available for CloudHub runtime planes?
No, but for Fabric

What is the difference between Edge policies and API policies?

Policies are applied in the Edge gateway or in the API Implementation / API proxy app
Edge policies apply to many API instances, API policies apply to exactly ONE API instance
Different policies are available

Where is Anypoint Security and Tokenization usually deployed?

Edge usually in a DMZ in a customer-hosted environment
Token usually inside the firewall to replace sensitive data with face data

What is Tokenization?
It replaces sensitive data with fake data in the same format

Credit card numbers, social security numbers, phone numbers etc.

What are the three main features of Edge security?

CAP: Content Security

Prevents malicious content from reaching a service during runtime execution
Pattern Scans
SQL injections
Limit message size
Limit number of file attachments
Service will generate a CAP violation if the request matches any of the above criteria

QoS: Quality of Service

Request data rate
Raw request data rate
Response data rate
Failed response rate
Message buffer utilization

DoS: Denial of Service

Alerts
Block-intervals: rejects messages
Block-forever: Permanently block messages
Shape-interval: Restricts the rate
Shape-forever: Restricts the rate forever

How does a symmetric key works?

One key is used to both encrypt and decrypt the information
How can you securely exchange a symmetric key over a public network?

How does an asymmetric key works?

Public key can be freely distributed since its only used to lock the data
The private key is used to decrypt the data

Alice to Bob: Hello lets set up a SSL session
Bob to Alice: Here is my Certification
Alice to Bob: Here is the one time encryption key for the session
Bob decrypts session using the private key and established the secure session

The certification must be valid
Certification must be signed by someone

What is the difference between:

Tokenization
Encryption
Data Masking
Token: Data to random String, You can not derive the original data from the token
Encryption: You can decrypt it again
Data Masking: One way process, you can not get the original value back

How can you secure application properties?

Create a Secure Properties Config
(It supports YAML configuration and Java properties)

The encryption can be given when starting runtime

Encrypted value is defined in ![value] in quotes
You can access it in the properties: ${secure::db.port}

How can you hide secure properties on cloudhub runtime manager?

Add it to the mule-artifact.json
The secure properties on cloudhub will be *

How can you secure data in Mule applications?

Mule cryptography module supports Java Cryptographic architecture:

symmetric and asymmetric encrypting/decrypting
Signing and signature validation of a payload or part of a payload

DataWeave also has a Crypto module for hashing payload:

How does Crypto Module works?

Java Keytool generate keystore file
Mule Crypto processor use JCE configuration with generated key store file

What is contained in a public certificate?

Public key
Organization details
The certificate issuer
A certificate is typically signed by a trusted certificate authority = CA or can be self signed using the Java Keytool

What is a Java keystore and a truststore?

Keystore stores public certificates plus corresponding private keys

The Java keytool can be used to create keystores for Mule applications or Anypoint Platform

Truststore stores public certificates for other parties and maybe also the signing of CAs

How do you configure one-way TLS for HTTPS?

How do you configure a two-way TLS for HTTPS?

Client sends identity: public certificate containing public key
Client has its own keystore for the private key
Server has its own truststore to validate the clients certificate

How can you externalize a Mule runtimes certificate?
On customer-hosted Mule runtime:

You can externalize to a folder outside the Mule_HOME location

You can also use external Secret Manager (Jenkins etc)

What is a Virtual Private Cloud = VPC?

Anypoint can create a private isolated network in the cloud to host CloudHub workers
Mule applications deployed to the VPC can communicate with each other using VPCs private network addresses

Where can you set a VPC with a DLBs?
VPC can remain completely isolated from external networks:

Usually a VPC uses one or more dedicated load balancers to route traffic with external public and private networks
A DLB can handle Mule event payloads up to 200MB

How can you connect VPCs?

Configure customized firewall rules for Mule applications deployed into the VPC:

Open ports beside 80, 8081, 8082
Block direct connections to CloudHub workers over port 8081 or 8082
Allow or restrict any other TCP or UDP ports

Connect a VPC to a corporate intranet

On-Prem or Clouds
Via VPM connection to a private network

Connect your VPC to another AWS VPC using VPC Peering

How can you hide DLB hostnames?

A DLB DNS entries are A records and are maintained by Mulesoft
Customers can define their own domain names as CNAMEs in their own DNS server

What is important about the VPC sizing?

Mule applications are deployed in CloudHub workers and each worker is assigned with a dedicated IP
For zero downtime deployment each worker in CloudHub needs additional IP addresses
A few IPs in a VPC are reserved for infrastructure
The IP addresses are usually in a private range with a subnet block specifier such as 10.0.0.1/24

What is the smallest and largest CIDR network subnet block you can assign for your VPC?
CIDR = Classless interdomain routing

smallest: /24 = 256 IP addresses

largest: /16 = I 65536 IP addresses

If you have:

4 different environments: Dev, staging, performance and production
50 application deployed in each environment
Two separate VPCs for production and non-production environments
In performance and production environment each Mule application will be deployed to two workers

What is the minimum CIDR network subnet for each of the VPCs? (non-prod and prod)
IP addresses

50 dev
50 staging

50 + 50 performance –> due to 2 workers per application

50 + 50 prod –> due to 2 workers per application

300 IPs + zero downtime 50% (150) = 450

/23 = 510 IPs so you select /23

Remember you can configure a VPC with any CIDR subnet between /24 and /16 without any additional costs

You should plan for worst case growth

Related Posts