DevOps Torino Meetup - SRE Concepts

GOOGLE SRE
CONCEPTSTorino DevOps Meetup

FROM THE KICK-OFF
PRESENTATION

Is SRE an alternative to DevOps?
Google creates and evolved the concept
independently from DevOps movement.
«If you think of DevOps like an interface in a programming language,
class SRE implements DevOps. SRE includes additional practices and
recommendations that are not necessarily part of the DevOps
interface. DevOps and SRE are not two competing methods for
software development and operations, but rather close friends
designed to break down organizational barriers to deliver better
software faster.
DevOps emerged as a culture and a set of practices
that aims to reduce the gaps between software
development and software operation.
The DevOps movement does not explicitly define how
to succeed.
SRE prescribes how to succeed in the various DevOps
areas.” (Liz Fong-Jones, Seth Vargo)
DEVOPS VS SRE (SITE
RELIABILITY ENGINEERING)

DEVOPS MANIFESTO –
SRE MANIFESTO

Site reliability engineering (SRE) was born at Google in 2003, prior to the
DevOps movement.
“SRE is what happens when you ask a software engineer to design an
operations team.” - Ben Traynor, VP of engineering at Google and founder of
Google SRE
“SRE is fundamentally doing work that has historically been done by an
operations team but using engineers with software expertise and banking on
the fact that these engineers are inherently both predisposed to, and have
the ability to, substitute automation for human labour. In general, an SRE
team is responsible for availability, latency, performance, efficiency, change
management, monitoring, emergency response, and capacity planning.” -
Ben Traynor, VP of engineering at Google and founder of Google SRE
Site reliability engineers create a bridge between development and
operations by applying a software engineering mindset to system
administration topics.
They split their time between operations/on-call duties and developing
systems and software that help increase site reliability and performance.
Google puts a lot of emphasis on SREs not spending more than 50% of their
time on operations and considers any violation of this rule a sign of system
ill-health.
The ideal site reliability engineer candidate is either a software engineer with
a good administration background or a highly skilled system administrator
with knowledge of coding and automation.
“SRE teams are characterized by both rapid innovation and a large
acceptance of Change” - Ben Traynor, VP of engineering at Google and
founder of Google SRE
THE ORIGIN
Ben Traynor, VP of engineering at Google

“Software engineering has this in common with having
children: the labour before the birth is painful and difficult, but
the labour after the birth is where you actually spend most of
your effort. Yet software engineering as a discipline spends
much more time talking about the first period as opposed to
the second [omissis] there must be another discipline that
focuses on the whole lifecycle of software objects, from
inception, through deployment and operation, refinement,
and eventual peaceful decommissioning.”
“Running a service with a team that relies on manual
intervention for both change management and event handling
becomes expensive as the service and/or traffic to the service
grows”
“Traditional operations teams and their counterparts in
product development often end up in conflict, most visibly
over how quickly software can be released to production. At
their core, the development teams want to launch new
features and see them adopted by users. At their core, the ops
teams want to make sure the service doesn’t break while they
are holding the pager. Because most outages are caused by
some kind of change a new configuration, a new feature
launch, or a new type of user traffic the two teams’ goals are
fundamentally in tension.”
THE PROBLEM TO
SOLVE

Site : the terms site comes from the original duty of
the newly born role, having focus on google.com web
site.
Reliability: “reliability is the most fundamental
feature of any product: a system isn’t very useful if
nobody can use it!” (Ben Traynor)
Engineering: “we apply the principles of computer
science and engineering to the design and
development of computing systems” “an SRE team
must spend 50% of its time actually doing
development.” (Ben Traynor)
.
THE NAME

GOOGLE SRE – CONCEPTS –
SRE VS DEVOPS

DevOps movement in the community and SRE initiative
in Google started from the same problem, the inefficiency
of having Developers and Operators working on the
different side of a wall, the first looking for feature and the
second for stability.
“One could view DevOps as a generalization of several
core SRE principles to a wider range of organizations,
management structures, and personnel. One could
equivalently view SRE as a specific implementation of
DevOps with some idiosyncratic extensions.“(Ben Traynor)
A DevOps Engineer is someone who understands the full
SDLC (Software Development Life Cycle)
DevOps focuses more on the automation part
SREs focus is more on the aspects like system availability,
observability, and scale
”The basic tenet of SRE is that doing operations well is a
software problem”
SRE VS DEVOPS

Reduce organisation silos: SREs share the ownership
of production with development teams and use the
same tools
Accept failure as normal: SRE’s concept of Error
Budget, avoiding 100% SLO
Implement gradual change: SRE is prescriptive about
usage of Canary Deployment
Leverage tooling and automation: SRE concept of
Toil and the guideline of “automate this year’s job”
Measure everything: SRE is prescriptive about
measuring Error Budget and Toil
SRE VS DEVOPS

DEVOPS VS. SRE: COMPETING STANDARDS OR FRIENDS?
(CLOUD NEXT ‘19) (@SETHVARGO)

Service Level Indicators (SLIs): metrics over time such
as request latency, throughput of requests per
second, or failures per request
Service Level Objectives (SLOs): targets for the
cumulative success of SLIs over a window of time
Service Level Agreements (SLAs): a promise by a
service provider, to a service consumer, about the
availability of a service
SLI, SLO, SLA – A
DEFINITION

“In general, for any software service or system, 100% is
not the right reliability target because no user can tell the
difference between a system being 100% available and
99.999% available.”
100% is not a viable availability target. Having a 100%
availability requirement severely limits a team or
developer’s ability to deliver updates and improvements
to a system.
To set the target isn’t a technical question at all, it’s a
product question, the business lead and the product
management must establish the system’s availability
target:
• What level of availability will the users be happy with,
given how they use the product?
• What alternatives are available to users who are
dissatisfied with the product’s availability?
• What happens to users’ usage of the product at
different availability levels?
SLO - TARGET

SLIs must be selected on the base of business value considerations.
Valuable SLIs:
• Request latency
• Batch throughput
• Failures per request (error rate)
Non valuable SLIs:
• CPU Time
• Memory Usage
• Operating System Uptime
Metrics must be aggregated overtime (for example “last 5 minutes”)
and a function as a percentile to be applied (for example “99
percentile”)
Metrics must be defined in advance and be known and accepted
across all organisation
Measurement must be reliable and not only the definition, also the
implementation of the measurement must be clearly defined and
approved
SLI - SELECTION

ERROR BUDGET

Error budget is a quantitative
measurement to establish the ratio
between work on implementing new
features and work on improving stability
“The error budget provides a clear,
objective metric that determines how
unreliable the service is allowed to be
within a single quarter. This metric
removes the politics from negotiations
between the SREs and the product
developers when deciding how much
risk to allow”
ERROR BUDGET AND
RISK

“If SLO violations occur frequently
enough to expend the error budget,
releases are temporarily halted while
additional resources are invested in
system testing and development to
make the system more resilient,
improve its performance, and so on.”
ERROR BUDGET AND
RISK

Use measurement to calculate expected
cost in terms of error budget of each
risk
Prioritise work on improvement on the
base of the computed impact of each
risk in terms of error budget
Cost for mitigating the risks could
require a review of the SLO due to
business value considerations
RISK MANAGEMENT AND
PRIORITISATION

GOOGLE SRE – CONCEPTS
– TOIL

Toil is not simply "work I don't like to do.”
Toil is not overhead (commuting, expenses reports,
meetings, …)
Toil is specifically tied to the running of a production
service.
It is work that tends to be manual, repetitive,
automatable, tactical and devoid of long-term value.
Toil is also reactive, as intervention done due to an alert.
When SREs find tasks that can be automated, they work to
engineer a solution to prevent that toil in the future.
Toil is not always bad. Predictable, repetitive tasks are
great ways to onboard a new team member and often
produce an immediate sense of accomplishment and
satisfaction with low risk and low stress.
TOIL AND TOIL
BUDGET

“We ensure that the teams consistently spending less than
50% of their time on development work change their
practices. Often this means shifting some of the
operations burden back to the development team, or
adding staff to the team without assigning that team
additional operational responsibilities.”
SREs must not be spending more than 50% of their time
on supportoperation activities
Developers should be involved in support regularly, but
their involvement become mandatory if a product
requires SREs to spend more than 50% of time on
operations
At least eight people need to be part of the on-call team to
correctly balance the load
Each on call person must handle no more than two events
per on-call shift to assure right quality
To preserve readiness for an effective on call support
practice handling hypothetical outages
ON CALL SUPPORT

“If you currently assign tickets randomly to victims on
your team, stop . Doing so is extremely disrespectful
of your team’s time, and works completely counter to
the principle of not being interruptible as much as
possible. Tickets should be a full-time role, for an
amount of time that’s manageable for a person. If you
happen to be in the unenviable position of having
more tickets than can be closed by the primary and
secondary on-call engineers combined, then structure
your ticket rotation to have two people handling
tickets at any given time. Don’t spread the load across
the entire team.”
SUPPORT TICKET

OBSERVABILITY AND MONITORING

Each monitoring system should address
two questions: what’s broken
(symptom) and why (cause)
Two types of monitoring can be defined:
• White-box monitoring: it inspects the
internal state of the target service
(application components metrics,
traces, logs). Focus on causes.
• Black-box monitoring: it accesses the
systems from external, as a real user
(httptcp probes, dns resolution,
network ping). Symptom-oriented.
Active recognition of error condition.
WHITE-BOX AND BLACK-
BOX MONITORING

Monitoring may have only three output types:
• Pages - A human must do something now
• Tickets - A human must do something within a few
days
• Logging - No one need look at this output
immediately, but it’s available for later analysis if
needed
”Putting alerts into email and hoping that someone
will read all of them and notice the important ones is
the moral equivalent of piping them to /dev/null :
they will eventually be ignored.”
An alert must be analysed, if an alert is ignored,
remove the alerting rule.
MONITORING AND
ALERTING

Collecting metrics and alerting are
business driven activities.
Collecting metrics and manage alerting
are different thing. You need to collect
metrics to have visibility on your
systems, but you do not have to use all
metrics to generate alerts.
Measurement, monitoring and alerting
must be related to SLOs.
Do not alert on anything, alerts must
report a service impact and be
actionable.
OBSERVABILITY

Latency
The time it takes to service a request. It’s important to
distinguish between the latency of successful requests and the
latency of failed requests. For example, an HTTP 500 error
triggered due to loss of connection to a database or other critical
backend might be served very quickly; however, as an HTTP 500
error indicates a failed request, factoring 500s into your overall
latency might result in misleading calculations. On the other
hand, a slow error is even worse than a fast error! Therefore, it’s
important to track error latency, as opposed to just filtering out
errors.
Traffic
A measure of how much demand is being placed on your system,
measured in a high-level system-specific metric. For a web
service, this measurement is usually HTTP requests per second,
perhaps broken out by the nature of the requests (e.g., static
versus dynamic content). For an audio streaming system, this
measurement might focus on network I/O rate or concurrent
sessions. For a key-value storage system, this measurement
might be transactions and retrievals per second.
THE FOUR GOLDEN
SIGNALS (1/2)

Errors
The rate of requests that fail, either explicitly (e.g., HTTP 500s),
implicitly (for example, an HTTP 200 success response, but
coupled with the wrong content), or by policy (for example, "If
you committed to one-second response times, any request over
one second is an error"). Where protocol response codes are
insufficient to express all failure conditions, secondary (internal)
protocols may be necessary to track partial failure modes.
Monitoring these cases can be drastically different: catching
HTTP 500s at your load balancer can do a decent job of catching
all completely failed requests, while only end-to-end system
tests can detect that you’re serving the wrong content.
Saturation
How "full" your service is. A measure of your system fraction,
emphasizing the resources that are most constrained (e.g., in a
memory-constrained system, show memory; in an I/O-
constrained system, show I/O). Note that many systems degrade
in performance before they achieve 100% utilization, so having a
utilization target is essential.
THE FOUR GOLDEN
SIGNALS (2/2)

INCIDENT MANAGEMENT

Effective incident management is key to limiting the
disruption caused by an incident and restoring normal
business operations as quickly as possible.
A well-designed incident management process has
the following features:
• Recursive separation of responsibilities
• Incident command
• Operational work
• Communication
• Planning
• A recognised command post
• Live incident state document
• Clear handoff
INCIDENT
MANAGEMENT

”It is better to declare an incident early and then find
a simple fix and close out the incident than to have to
spin up the incident management framework hours
into a burgeoning problem.”
If any of the following is true, the event is an incident:
• Do you need to involve a second team in fixing the
problem?
• Is the outage visible to customers?
• Is the issue unsolved even after an hour’s
concentrated analysis?.
WHEN TO DECLARE AN
INCIDENT

Incident Management Roles:
• Incident Commander
• The incident commander holds the high-level state about the
incident. They structure the incident response task force,
assigning responsibilities according to need and priority.
• Operations lead
• The Ops lead works with the incident commander to respond
to the incident by applying operational tools to the task at
hand. The operations team should be the only group
modifying the system during an incident.
• Communication lead
• This person is the public face of the incident response task
force.
• Planning lead
• The planning role supports Ops by dealing with longer-term
issues, such as filing bugs, arranging handoffs, and tracking
how the system has diverged from the norm so it can be
reverted once the incident is resolved.
• Logistics lead
• The logistic role supports Ops by dealing with things as
ordering dinner or dealing with vendors for spare parts.
INCIDENT
MANAGEMENT

POSTMORTEMS

”Postmortems should be blameless and
focus on process and technology, not
people. Assume the people involved in
an incident are intelligent, are well
intentioned, and were making the best
choices they could given the
information they had available at the
time.”
Questions about which data where
available, how the system was
behaving, which actions done and their
effect. Avoid questions about why an
action has been done or why not.
POSTMORTEMS AND
RETROSPECTIVES

Content of a postmortem:
• Incident summary
• Detailed timeline
• Detection
• Impact
• Root Causes
• Triggers
• Mitigation and Resolution
• Lesson learned
• What went well
• What went wrong
• Where we got lucky
• Action Items (with clear owner)
POSTMORTEMS AND
RETROSPECTIVES

ENGAGEMENT MODEL

Not all Google services receive close SRE
engagement.
Google defines three different
engagement models
• Simple PRR (Product Readiness
Reviews) Model
• Early Engagement Model
• Frameworks and SRE Platform
SRE ENGAGEMENT
MODEL

Development team requests that SRE take over
production management of a service, one to three SREs
are selected to conduct the PRR process.
The discussion covers matters such as:
• Establishing an SLO/SLA for the service
• Planning for potentially disruptive design changes
required to improve reliability
• Planning and training schedules
The Training phase unblocks onboarding of the service by
the SRE team. It involves a progressive transfer of
responsibilities and ownership of various production
aspects of the service, including parts of operations, the
change management process, access rights, and so forth.
To complete the transition, the development team must
be available to back up and advise the SRE team for a
period of time as it settles in managing production for
the service. This relationship becomes the basis for the
ongoing work between the teams.
SIMPLE PRR MODE

The Early Engagement Model introduces
SRE earlier in the development lifecycle.
SRE participates in Design and later
phases, eventually taking over the
service any time during or after the
Build phase.
This model is based on active
collaboration between the development
and SRE teams.
EARLY ENGAGEMENT
MODEL

SRE builds framework modules to implement canonical
solutions for the concerned production area. As a result,
development teams can focus on the business logic,
because the framework already takes care of correct
infrastructure use. A framework essentially is a
prescriptive implementation for using a set of software
components and a canonical way of combining these
components.
The service frameworks implement infrastructure code
in a standardized fashion and address various production
concerns. Each concern is encapsulated in one or more
framework modules, each of which provides a cohesive
solution for a problem domain or infrastructure
dependency.
Framework modules address the various SRE concerns
enumerated earlier, such as:
• Instrumentation and metrics
• Request logging
• Control systems involving traffic and load
management
FRAMEWORKS AND
SRE PLATFORM

How can I know if I’m doing DevOps?
• Do I consider Developers, Test Engineers and Sys Admins as
members of the same team having one common goal?
• Do I know and care about end users features and
functionalities (business value)?
• Do I automate infrastructure and code deployment? Do I
understand and manage SDLC?
• Do I focus on collecting as much metrics as possible from
build process to production runtime?
How can I know if I’m doing SRE?
• Do I protect my 50% of the time for engineering activities?
• Do I define SLO and manage Error Budget?
• Do I continuously work to improve self-healing of
production solutions? Do I target “autonomous” solutions
and not only “automated”?
• Do I use a structured Incident Management process with
clearly defined roles?
AM I DOING SRE OR
DEVOPS?

DevOps Torino Meetup - SRE Concepts

Recommended

More Related Content

What's hot (20)

Similar to DevOps Torino Meetup - SRE Concepts (20)

More from Rauno De Pasquale (13)

Recently uploaded (19)

DevOps Torino Meetup - SRE Concepts

Editor's Notes