Xenit diary dev con 2018

ALFRESCO
DEVCON
2018
LISBON, PORTUGAL
JANUARY 16-18, 2018
DEVCON 2018
Inclusive and inspiring
topics from Alfresco
community.
INSPIRATION AT
WORK
How DevCon inspired
Xenit, by Ronny
Timmermans, CTO and
Founder
XENIT' TALKS
Five talks from Xenit
team, showing all our
expertise, as Alfresco
certified partner.
XENIT
DIARY
ALFRESCO
DEVCON
2018

"VERY WELL BALANCED
BETWEEN PURE
TECHNICAL AND
BROADER CONTENT"
SCAN THE QR CODE TO FIND
ALL PRESENTATIONS
FROM DEVCON
Peak of Mt. Sifonn, Bulgaria
Alfresco DevCon 2018
DEVCON IS
BACK
Last month, developers, engineers,
marketing and sales, customers,
enthusiasts and evangelists, they
were all together in Lisbon,
Portugal, attending the Alfresco
DevCon 2018.
LEARN.
CONNECT.
COLLABORATE
DevCon 2018 is an
international developer
conference entirely dedicated
to Alfresco technology. With
the support of the
community, customers, and
partners, DevCon, you will
increase your technical know-
how, connect you with other
Alfresco developers, and let
you collaborate with our
team and each other.
"

Peak of Mt. Sifonn, Bulgaria
WHO WILL BE THE
NEXT JEDI?
You may be surprise of this
association with Star Wars, but it
was the theme used by John
Newton, Alfresco CTO and
Founder, to provide his vision on
the next disruptive technologies.
In order:
• Artificial intelligence
• Microservices Architecture
• Blockchain
• Cloud operating System
• Edge Computing
• Multi-model User interface
The hearth of the Content
Service is intelligence and in the
next future a set of technologies
(NLP, machine learning) will be
one gaming – changing for a
deeper and wider understanding
of the information.
CLOUD IS THE NEW
NORMAL
Production ready docker
stackDevCon is not the first
event where it has been
mentioned, but now it’s quite
clear that the platform is
preparing to move to the
Cloud.
Amazon Web Service Cloud
Service, precisely.
As Angel Borroy highlights in
his blog post, "it seems clear
from both talks, that Alfresco
has started an isolation
services process to prepare
the platform to be deployed
natively on Cloud. Probably,
Activiti 7 (not available yet)
will be the first product to be
transformed in that direction,
so we have to wait a little to
start playing with these
technologies."
ADF AND THE
FUTURE OF
ALFRESCO SHARE
This topic created a lot of
noise in the Alfresco
community.
During the Q&A session,
Thomas DeMeo, VP of the
Product Management,
answered a question about
the future of Alfresco Share,
saying the Share will be
deprecated in the upcoming
years, and ADF
(APPLICATION DEVELOPMENT
FRAMEWORK) will fill in a
certain way this gap
In the Jeff Potts’ article, you
can read the effect of this
decision and what might
happen next...

INSPIRATION
AT
WORK
Ronny Timmermans
CTO and Founder @Xenit

Devon 2018 inspired Xenit to help
re-inventing Content Services in
next decade. "ECM is not dead, Jim
but it is transforming".
Cloud and A.I. are creating new
ways to manage content and
automate the processes around it.

The cloud creates and advances
future application architectures,
even on premise, and monoliths
are even more outdated than
Dinosaurs
We retained a number of striking
topics at DevCon 2018, in line with
our sense of purpose to
bring Cognitive Clarity @Zero
latency for different markets,
verticals and horizontal challenges
like GDPR:
1. Alfresco is heading towards a
new micro-services architecture
and Xenit actually has built a
number of components to
accelerate that journey (Alfred Edge
& Inflow).
Think docker containers and
kubernetes, even for small scale
solutions.
2. In the future, a Gateway will be
part of the Alfresco reference
architecture. We are looking
forward to build this with, for or
around Alfresco.

3. AI has a bunch of use cases in
content and process services. Edge
computing, Cloud, Blockchain, multi-
modal interfaces and related
technologies will make their mark in
future content and process services.
4. Multi-modal *cognitive*
interfaces will be the new norm,
application and business integration
will be smart and dynamic
5. Graphql, as launched by
Facebook, fitting within our react-
stack, is a very interesting
specification for query and
modification (mutation) and
notification. It is used for interfacing
with Activiti.
6. Alfresco is thinking of using
Electron for packaging ADF
applications on a desktop.
"Xenit seeks to
develop these best
practices, functionality and
automation into our Alfred
product line, with Alfresco
technology at its core".
Ronny Timmermans

At Xenit, we think we can help
Alfresco and its partners advance
on this journey by :
Ronny Timmermans
1. Creating a community
around Facebook's react and
extending the ADF reach into
the react community,
2. Extending our intelligent
gateway Alfred Edge for
synchronous and
asynchronous
communication, while
infusing best 'edge
computing' practices into it.
3. Continuing to deliver
simple, smart and swift
windows desktop experience
with Alfred Desktop
COGNITIVE INTERFACE
INTELLIGENT INTEGRATION
AND EDGE COMPUTING

XENIT TALKS
Roxana Angheluta
Thijs Lemmens
Jasper Hilven
Willem Van den Eynde
Younes Regaieg

PRODUCTION READY
DOCKER STACK
Roxana Angheluta
Thijs Lemmens
Jasper Hilven
Younes Regaieg

Roxana Angheluta: "Production ready Docker stack"
THE CHALLENGES
Automate the process of
installing and maintaining
Alfresco.
Have Alfresco image(s) prepared
ready, both for the Community
and Enterprise, with or without
Solr and LibreOffice running
inside, for clustered
environments as well as non-
clustered, remotely or on our
hosted platform - reusing
services such as the smtp
server, with easy configuration
changes and easy upgrades.
Keep track of all changes
involved and install a full
monitoring, which alerts in case
of problems and helps in
capacity planning for resources.
Flavor
(community vs
enterprise)
Automation
Upgrades
Monitoring
Traceability Versioning
Single
deployment
method
Easy
configuration
changes
Data separation Security
Alerting
Single:
Bundled /
separated
HostedClustered

THE ARCHITECTURE
1. Docker-compose up/down to
start / stop services.
2. Consul (Hashicorp) + registrator
(GliderLabs) for
service discovery.
3. Consul template (Hashicorp) for
rendering.
"We started simple, with docker-compose files to start and stop services. We use Consul and registrator for service discovery and Consul
template for rendering load balancing configurations and restarting load balancer. Consul was developed by Hashicorp and is a key-value
store for storing services, adding or removing them on the fly based on health checks. The actual registration is done by registrator
developed at GliderLabs." - Roxana Angheluta, Senior ECM Engineer @Xenit
We already deployed this architecture to our clients, with 100% successful use cases

THE DOCKER IMAGES
The team started running the
installer on a base Ubuntu
image, very simple to implement
and to upgrade to higher
Alfresco versions.
When more flexibility is needed,
like a specific version of java /
tomcat, Installer is not right
choice. The team moved
therefore to a different method
to build images.
As base layer they started from
java, on top of which they added
Tomcat, then a skeleton for an
Alfresco including the init script,
then the war files. Finally, client
specific modules were added on
top.
Watch the talk
We aim for flexibility in customizing. We make the container stateless, by mounting the
data and the logs.

THE MONITORING SYSTEM
Our monitoring stack is
complex and includes a
number of parts:
• Kibana, for log monitoring
• Grafana, for numerical data
• Cabot for alerting
• Alerta which aggregates
Cabot warnings and alerts
and allows you to spot and
prioritize issues, across
multiple clients.
What's next? Our next challenge is using reporting data to automatically produce
SLA (Service Level Agreement) reports.

PostgreSQL FOR ALFRESCO:
THE PRACTICAL GUIDE
Roxana Angheluta
Thijs Lemmens
Jasper Hilven
Younes Regaieg

Thijs Lemmens: "PostgreSQL for Alfresco: the practical guide"
During DevCon, Thijs Lemmens, Senior ECM Engineer @Xenit, focused on three PostgreSQL topics:
Master -Slave
replication
Continuous backup
and Point-in-time
recovery
Upgrading
database without
downtime

"Synchronous or Asynchronous replication. That is the question"
In PostgreSQL, a master-slaves setup relies on a concept called streaming replication. This concept relies on the
transaction logs (WAL) that ensure ACID compliance and copies those over from master to slave,or standby.
The advantage of synchronous replication is that when the master returns from a commit, you are sure that the
commit is also safely written to the slave. In case of a disaster on the master, you don’t loose any commits.
However, when you want high availability, you need to have at least 2 slaves. Otherwise, the master will block
writes, when the slave is down. Synchronous replication can have a large impact on performance, because it is
highly dependent on the network latency, and throughput.
When using asynchronous replication, you don’t have these drawbacks, but you have to take into account that
there will be a replication lag. It is important to monitor this lag and alarm when the lag is higher than what you
think is acceptable for your use case. In case of disaster, you might loose some transactions. This also means that
you will have to restart Alfresco, because you risk having in consistent caches otherwise.

CONFIGURATION FOR DEMO PURPOSES
postgresql.conf:
wal_level = hot_standby # or higher
max_wal_senders = 20
pg_hba.conf:
host replication repuser 172.18.0.0
255.255.255.0 md5
pg_basebackup --pgdata=datadir --xlog-
method=stream --write-recovery-conf --
progress --dbname="host=pgmaster
port=5432 user=repuser
password=pwd"
Conguring the slave is very simple. You
just need to run a command, and
everything will be copied from the
master, including the conguration
Alfresco uses a JDBC connection
string. You can dene multiple
servers in a PostgreSQL JDBC
connection string:
jdbc:postgresql://pgmaster:5432,pgs
lave:5432/alfresco
To congure PostgreSQL for streaming replication we have to adapt 2 les in the PostgreSQL data directory:
1. postgresql.conf - this is the main conguration le.
2. pg_hba.conf - In this le we can congure who as access to what database, from where.
MASTER SLAVE ALFRESCO

CONTINUOUS BACKUP AND POINT-IN-TIME RECOVERY
Wal archiving concept: the wal files are copied to another location. The copied wal files can then be used
when we want to restore a backup. This method has (at least) 3 advantages:
1. You don’t need to schedule a nightly backup.
2. The backup is done continuously
3. Upon restore, you can decide to restore up until a certain time or transaction. If you made an error, like
accidentally dropping a table, you can restore just before you made the mistake.
WAL ARCHIVING
POINT-IN-TIME RECOVERY
archive_mode = on
archive_command = 'test ! -f /wal_archive/%f && cp %p /wal_archive/%f'
To start the wal archiving we need to add the following lines in the postgresql.conf:
Before starting the recovery, we need to restore the last base backup, taken before the time we want to
recover. This base backup should become the new data directory. In that new data directory you need to
put a recovery.conf:
The archive_commandparameter uses 2arguments:
• %f: The name of thewal file
• %p: The path of thewal file to copy
restore_command = 'cp /wal_archive/%f %p'
recovery_target_time = '2018-01-09 09:39:50'

UPGRADING WITHOUT DOWNTIME
The easiest way to perform a
database upgrade is using
pg_dump to make a logical backup
and to restore that backup on the
new version of the database. The
problem with this approach is that
you need to perform these
operations, and your database
should be down, or in a read-only
state.
A classic master-slave replication
looks like viable solution. This way
you could add a slave of the new
version and switch over when all
transactions are replicated.
However, because the replication is based on a binary format, that changes between version, you can only use
physical streaming replication between the same versions of PostgreSQL.
With the help of an extension pglogical, we can overcome this problem. Pglogical translates the physical
statements to a version independent format, so they can be applied on another version of PostgreSQL.

HOW TO GET THINGS
DONE FAST!
Roxana Angheluta
Thijs Lemmens
Jasper Hilven
Younes Regaieg

Jasper Hilven: "How to get things done fast"
Alfred Finder is a web application that allows you to quickly find documents on an Alfresco back-end, preview
them and process metadata.
When customers use a software, most of the time, they don't think about the choices that are behind the
development of it.
In this talk, Jasper Hilven, Software Engineer @Xenit, focused on two many aspects of building software:
1. Managing complexity and
2. Ensuring quality at multiple levels.
BEHIND THE SCENES OF ALFRED FINDER

MANAGING COMPLEXITY TO KEEP THINGS SIMPLE
Using React is a logic
consequence of keeping
things simple
React gives you the power
of easily building the UI of
your application without
doing other complex things
that we do not need.
It just works
In progress Review Done
As software is build out of
components, keep our
components as simple as
possible.
This implies that the
complexity of a software
component should grow
naturally. This means you
start with a simple value.
As soon as you want to
reuse the value, you make
a variable. If your variable
needs a scope, you build a
function or even later a
separate class.
We have multiple projects
to ensure the right
granularity of reuse, but
also because we have
multiple customers.
Customers are also not
happy if you deploy their
custom code to other
customers.
We use NPM to manage the
dependencies between
these projects.

ENSURE QUALITY AT MULTIPLE LEVELS
You need to find a balance between getting things done and assuring quality.
Quality of the code
@Xenit, we use Typescript to
typecheck each single line of code.
The cost of these checks are
rather low, and happen each time
you write a single line of code. We
also use a Linter for style
checking. Its called Typescript
Linter. This is very useful if you
have multiple people working on
the same project.
Components are the building block
of software. We use unit tests to test
our components. As soon as
somebody breaks a component, a
test complains about it. Then we can
fix the issue long before the
customer notices anything. "Its like
having a machine that checks all the
bricks each time again and again,
before you are even building your
house."
We automatically run (almost)
everything a user can do, each time
we change something to product. This
ensures that the customer can do with
our software what we promise.
Jenkins runs all these tests and keeps
all the reports of all the performed
tests.
Quality of the components Quality of checks

DON'T REINVENT THE
WHEEL. REUSE!
Roxana Angheluta
Thijs Lemmens
Jasper Hilven
Younes Regaieg

Younes Regaieg: "Don't reinvent the wheel. Reuse!"
Most software / systems are
actually designed by orchestrating
other smaller components (or
bricks) that have been crafted or
at least used in a different context.
Software engineering was initially
more focused on original
development but now it is
recognized that to achieve better
software faster and more
efficiently, a design approach
based on systematic reuse is
actually needed.
"Actually reuse of existing assets in the software industry is not as new as some might think.
Even though this industry started by promoting original software development, it definitely recognized that, in order to
reach a better, faster and prone proof software, one would need to make use of reusable assets and directs most of the
engineering efforts towards making the missing pieces in the puzzle, or maybe enhancing existing ones in order for them to
fit perfectly in their environment."
-Younes Regaieg, Senior Software Engineer-

WHAT CAN BE REUSED?
1. SOLUTIONS: For instance, a bunch of
docker container images / or VMs (Virtual
Machines) to speed up the deployment
process.
2. APPLICATIONS: By adopting Alfresco
Content Services, a fully fledged
applications are reused
3. COMPONENTS: A component can be
actually be an add-on or a library that
bundles a feature or set of tightly
coupled/related features.
4.OBJECTS AND FUNCTIONS: Same as
components, they represent a piece of
reusable logic, but at a smaller scale.

A SUSTAINABLE SOFTWARE REUSE PROCESS
IN FIVE STEPS
1. ANALYZE SPECIFICATIONS: Usually at the start of a project you get the software specifications which you use
to create a component breakdown to be addressed one piece at a time.
2. PICK ELEMENT: Then you start picking pieces and address them one by one.
3. FIND REUSABLE ASSETS: This step would highly depend on the knowledge of the team on board as well as
information that is available in the context of the project.
1 2 3 4.A
4.B
5

FIND REUSABLE ASSETS USE CASE
In this use case, we are looking for the
company internal assets catalogs for
reusable assets and optionally looking
within partners portfolios and
community projects.
This depends on the customer in
question, the required support level, the
available budget and the experience of
the team members involved.
There are two options:
First option: Not finding any assets that
could be useful, exit and terminate the
activity
Second option: Find assets and run
them though legal/sales departments

A SUSTAINABLE SOFTWARE REUSE PROCESS
IN FIVE STEPS
4.A IMPLEMENT NEW FEATURES: If you didn't find any assets, fit for the reuse, then you can implement the new
features
4.B. CONFIGURE AND EXTEND: Our search might have been a success and so we go forward to the step where
we configure or extend (or even implement something on top of it) the found re-usable assets.
5. MANAGE POTENTIAL ISSUES: You might eventually land on an asset where we can not cover all requirements
out of the box or we encounter a bug that needs resolving. in such case you might escalate to a new path where
patch the reusable asset and eventually send out a contribution for the project in question.
Following a couple of iterations on this process, your organization will start to develop a knowledge base for reusable
addons, with an increased productivity, faster time to market and definitely a better software quality
1 2 3 4.A
4.B
5

3 TIPS TO MIGRATE
ALFRESCO TO AWS OR
AZURE
Roxana Angheluta
Thijs Lemmens
Jasper Hilven
Younes Regaieg

Cloud infrastructure offers many benefits including cost-savings and transparency in architecture.
Willem Van den Eynde, Senior ECM Sofwtare Architect @Xenit, got in touch with AWS services in 2012.
However client adoptions was lacking behind. Until 2016, when he did two migrations from an existing on-
premise installation to AWS and a similar one to Microsoft Azure.
Willem Van den Eynde " 3 tips to migrate Alfresco to AWS or Azure

TIP #01 - COPY DATA FIRST: Through the 2 migration projects, copying over the data has been the bottleneck.
It's often slow to copy over lots of small files. To give an example: it takes about 15 days to copy over 4TB over a
3MPBps connection. Speeding up the process by using temporary storage solutions and copy over deltas also
adds to the overall project throughput.

TIP #02 - CLOUD ARCHITECTURE: Cloud architectures are very transparent, because all information about
their possibilities and prices are available online. This transparency helps to easy co-create the desired
architecture between client, Alfresco partner and infrastructure partner.
Components are regularly added and updated to take into account the latest technology and are easy to
evaluate and access.
.

TIP #03 - MIGRATE FIRST, THEN...: It's a migration project first. Those projects typically require a very short cut-
off period, during the weekend for example. Therefore it's best to limit the scope to just the migration. And
leave other ideas like upgrades to subsequent projects.

Xenit diary dev con 2018

More Related Content

Similar to Xenit diary dev con 2018 (20)

More from XeniT Solutions nv (20)

Recently uploaded (20)

Xenit diary dev con 2018