SlideShare una empresa de Scribd logo
1 de 38
Descargar para leer sin conexión
The great fire of 2021
Kris Buytaert
@krisbuytaert
@krisbuytaert
March 10 , 2021
●
03:59 Incoming Phone call
●
“Our Datacenter is On fire”
●
5 minutes of waking up..
going downstairs trying to
realize what was just said to
me, partial panic , partial .. it
can’t be THAT bad
@krisbuytaert
Fire
@krisbuytaert
Kris Buytaert
●
I used to be a Dev,
●
Then Became an Op
●
CTO and Open Source Consultant @inuits.eu
●
Everything is a freaking DNS Problem
●
Evangelizing devops
●
Organiser of #devopsdays, #cfgmgmtcamp, #loadays, ….
@krisbuytaert
Impact :
●
3.6 million websites
across 464.000
distinct domains
●
https://news.netcraft.com/archives/2021/03/10/ovh-fire.html
@krisbuytaert
Immediate Assesment ?
●
What Customer Facing Production services have we lost ?
●
What Internal Production services have we lost ?
– Do we need any of those to recover ?
●
What are the priorities ?
– Which services
●
24/7 money generating platforms
●
Office hours tooling
– What parts of these services
●
Basic Service, no redundancy
@krisbuytaert
Our Immediate visible Impact
●
Multiple customers failed over automatically to
the 2ndary DC
●
1 VIP address failed to move
●
1 platform was spread over SBG2 & SBG1
●
1 DR site destroyed
@krisbuytaert
Failing VIP
●
1 VIP address failed to move (API call failed)
●
Manual action to use a different public IP for the loadbalancer
●
Updated DNS records
●
Except for the 1 domain the customer controlled.
●
This cost us about 60 minutes before we could reach someone with
the right credentials to update the dns
●
Some customer devices had the VIP address hardcoded :(
– Manual interventions needed :(
@krisbuytaert
Customer Impact
●
1 platform was spread over SBG2 &
SBG1
●
Rebootstrapped the platform on a
different ISP, restored the backup
●
At 0900 am platform was ready for use,
●
We were improving performance of the
platform
●
Customer wouldn’t have known if we
didn’t tell them
@krisbuytaert
Total Loss:
●
We lost 13 physical
servers
●
135 vm’s
@krisbuytaert
Root Cause
●
Klaba does not say the UPS is the definite cause. "We don’t have all the answers today,"
he said. The OVHcloud staff responded to alarms at 11.42 pm on Tuesday, but the
affected part of the data center had already filled with smoke: "Two minutes after, they
took the decision to leave, because it was too dangerous."
●
The firefighter's thermal cameras found UPS7 and UPS8 on fire, but further data will be
extracted from on-site cameras: "We have 300 cameras in Strasbourg," said Klaba. "We
expect to have all the answers about how it started. We will give you all the information."
●
https://www.datacenterdynamics.com/en/news/ovh-fire-octave-klaba-says-ups-systems-
were-ablaze/
@krisbuytaert
Our Impact
NextCloud
●
Gitlab
●
Lots of development
environments
●
Lots of Test nodes
●
Partial K8s clusters
@krisbuytaert
We survived
●
18 hours of work
●
2 engineers at night
●
4-5 during daytime
●
Recovery
●
Communication
●
Restoring Backups
@krisbuytaert
At 0910
●
Another customer seemed to have issues
accessing some files
●
We had overlooked 1 gluster cluster that
seemed healthy but wasn’t
●
Switched to a single node mount point
(don’t try this at home)
@krisbuytaert
No data was lost
@krisbuytaert
Phase II
●
After 0900 we started restoring our internal tools from
backup
●
Verified all the secondary nodes would be making backups
●
Verified we had sufficient diskspace left
●
“Shoemaker always wears the worst shoes”
●
Challenge : Hardware !
@krisbuytaert
Hardware Landrush
●
Everybody needed new
hardware
●
We had spare hardware
ready for a new platform in
the wrong DC of a different
supplier . Temporary OK
●
We spun up new hardware
at a different supplier
@krisbuytaert
Wrong decisions
●
On day 2 we wasted at least 6 hours trying
spinning up a new supplier
●
Figuring out their network and redundancy
strategies was not the right time
●
We could scale on the existing suppliers
@krisbuytaert
Phase III
●
Restoring the rest of the services
●
What boxen would we be getting back ?
●
What boxen were foobar ?
●
(in the end we only got 3 physical servers back which were in SBG3),
the rest was lost
●
Priority on making sure all pipeline promotions of actively
developed platforms were running again
●
Hardware availability
@krisbuytaert
The day after
●
New DR platform where needed
●
Inventory of what besides production was impacted (lots of
development boxen, partial clusters)
●
Prioritization of what needs to be 100% back first
●
What hardware do we have / need additionally
@krisbuytaert
Your NEW DR Plan
●
You have just lost your DR site
●
You need to plan again
– Do we take the risk for now ?
– Do we respin a new DR ?
●
IAC + Datasync makes DR Trivial
@krisbuytaert
What saved us ?
●
Real Infrastructure as Code
●
Architecture
●
Backups
●
Fast escalation
@krisbuytaert
Real Infrastructure as Code
●
Desired State => Puppetize all the things
●
100% control , no 3rd parties involved
– Provision vm
– Deploy applications
– Deploy database schemas
●
No handovers , 1 person can deploy this, no other teams involved
●
Exported resources for Loadbalancers (haproxy), Monitoring (icinga), Databases (mysql)etc
– Heavy automation
●
All deployed versions are available in (yum) repositories
●
Everything is a Pipeline Puppet, Hiera, DNS, ... (jenkins)
@krisbuytaert
CloudNative vs Cloudnaive
●
We don’t own hardware
●
Baremetal on demand at $ < AWS
– Hetzner / ovh / ...
●
Spin up 120 seconds, decomission
when unneeded.
●
vm definitions are in code
●
Ansible to bootstrap, Puppet for
Desired state
@krisbuytaert
Multi Cloud
●
OVH
●
Hetzner
●
...
●
Workload is spread
●
Customer DR is in other
Supplier
@krisbuytaert
Multi Datacenter
●
OVH
– SBG, GRA, RBX, ..
●
Hetzner
– FSN X, HEL
@krisbuytaert
CloudNative ?
●
Critical Customer services are build redundantly,
(corosync, haproxy, mysql, elastic)
●
Losing a bare metal should doesnot have an impact
●
We can rebootstrap our nodes (kickstart+ puppet +
ansible)
●
All UGC data we know about is backed up (rsnapshot)
@krisbuytaert
Cloud Agnostic
●
A vm is a vm is a vm
●
Vendor Specific
Features
– e.g VIP addresses are
abstracted in code
@krisbuytaert
Dedicated stacks per Customer/Project
●
Single Purpose vm’s
●
Single Purpose Database Clusters
●
Single Purpose Storage Clusters
●
Single Purpose Loadbalancers
=> Reduced impact, Reduced complexity
@krisbuytaert
Clusters are Multi DC
●
Depending on scale , preference on HA vs
actual Loadbalancing
●
Pinning of resources within 1 DC
●
Lessons Learned: we need Multi Campus
@krisbuytaert
HA Applications
●
No local file usage
– If local files -> {gluster,drbd,...}
●
MySQL replication
●
Elastic multiple nodes (if we can’t generate the
data)
●
Multiple instances of rabbitmq
@krisbuytaert
HA Storage
●
VM’s on Local Disk
– 100% puppetized => no differences in config/ software
deployment
– “Disposable”
●
Storage distributed
– Gluster, DRBD, (Ceph)
●
(No Raid)
@krisbuytaert
Backups
●
Are both on site
●
And off Site
@krisbuytaert
Documentation
●
Mkdocs based
●
Git repo
●
Locally available
@krisbuytaert
Lessons Learned
●
Make sure you can always provision more resources on multiple
providers
●
A Datacenter is not always Datacenter, even with different buildings
you need Campus redundancy
●
Fix the shoemaker problem
●
Everything IS a fscking dns problem
– U WANT control over the DNS of the production sites you run
●
Or an MTU problem
@krisbuytaert
Thnx
Thnx to everyone who helped
out during the outage !
@krisbuytaert
March 10 , 2021
@krisbuytaert
Contact
Inuits
Inuits
Essensteenweg 31
Essensteenweg 31
Brasschaat
Brasschaat
Belgium
Belgium
891.514.231
891.514.231
+32 475 961221
+32 475 961221
Kris Buytaert Kris.Buytaert@inuits.eu
Kris Buytaert Kris.Buytaert@inuits.eu
Further Reading
Further Reading
@krisbuytaert
@krisbuytaert
http://www.krisbuytaert.be/blog/
http://www.krisbuytaert.be/blog/
https://inuits.eu/
https://inuits.eu/

Más contenido relacionado

La actualidad más candente

Repositories as Code
Repositories as CodeRepositories as Code
Repositories as CodeKris Buytaert
 
Deploying your SaaS stack OnPrem
Deploying your SaaS stack OnPremDeploying your SaaS stack OnPrem
Deploying your SaaS stack OnPremKris Buytaert
 
Continuous Infrastructure First
Continuous Infrastructure FirstContinuous Infrastructure First
Continuous Infrastructure FirstKris Buytaert
 
Continuous Delivery of (y)our infrastructure.
Continuous Delivery of (y)our infrastructure.Continuous Delivery of (y)our infrastructure.
Continuous Delivery of (y)our infrastructure.Kris Buytaert
 
Building and Deploying MediaSalsa, an Open Source DAM as Saas platform
Building and Deploying MediaSalsa, an Open Source DAM as Saas platformBuilding and Deploying MediaSalsa, an Open Source DAM as Saas platform
Building and Deploying MediaSalsa, an Open Source DAM as Saas platformKris Buytaert
 
ADDO 2019: Looking back at over 10 years of Devops
ADDO 2019:    Looking back at over 10 years of DevopsADDO 2019:    Looking back at over 10 years of Devops
ADDO 2019: Looking back at over 10 years of DevopsKris Buytaert
 
Can we fix dev-oops ?
Can we fix dev-oops ?Can we fix dev-oops ?
Can we fix dev-oops ?Kris Buytaert
 
Is there a Future for devops ?
Is there a Future for devops   ? Is there a Future for devops   ?
Is there a Future for devops ? Kris Buytaert
 
Pipeline as code for your infrastructure as Code
Pipeline as code for your infrastructure as CodePipeline as code for your infrastructure as Code
Pipeline as code for your infrastructure as CodeKris Buytaert
 
Dev secops opsec, devsec, devops ?
Dev secops opsec, devsec, devops ?Dev secops opsec, devsec, devops ?
Dev secops opsec, devsec, devops ?Kris Buytaert
 
From MonitoringSucks to Monitoring Love , 2016 Edition
From MonitoringSucks to Monitoring Love , 2016 EditionFrom MonitoringSucks to Monitoring Love , 2016 Edition
From MonitoringSucks to Monitoring Love , 2016 EditionKris Buytaert
 
Moby is killing your devops efforts
Moby is killing your devops effortsMoby is killing your devops efforts
Moby is killing your devops effortsKris Buytaert
 
Is there a future for devops ?
Is there a future for devops ?Is there a future for devops ?
Is there a future for devops ?Kris Buytaert
 
Velocity 2011: Production Begins in Development
Velocity 2011: Production Begins in DevelopmentVelocity 2011: Production Begins in Development
Velocity 2011: Production Begins in Developmentdev2ops
 
Automating MySQL operations with Puppet
Automating MySQL operations with PuppetAutomating MySQL operations with Puppet
Automating MySQL operations with PuppetKris Buytaert
 
Nightmare on Docker street
Nightmare on Docker streetNightmare on Docker street
Nightmare on Docker streetKris Buytaert
 
From devoops to devops
From devoops to devopsFrom devoops to devops
From devoops to devopsKris Buytaert
 
Devops is a Security Requirement
Devops is a Security RequirementDevops is a Security Requirement
Devops is a Security RequirementKris Buytaert
 
Devops is dead, Long Live Devops
Devops is dead, Long Live DevopsDevops is dead, Long Live Devops
Devops is dead, Long Live DevopsKris Buytaert
 

La actualidad más candente (20)

Repositories as Code
Repositories as CodeRepositories as Code
Repositories as Code
 
Deploying your SaaS stack OnPrem
Deploying your SaaS stack OnPremDeploying your SaaS stack OnPrem
Deploying your SaaS stack OnPrem
 
Continuous Infrastructure First
Continuous Infrastructure FirstContinuous Infrastructure First
Continuous Infrastructure First
 
Continuous Delivery of (y)our infrastructure.
Continuous Delivery of (y)our infrastructure.Continuous Delivery of (y)our infrastructure.
Continuous Delivery of (y)our infrastructure.
 
Building and Deploying MediaSalsa, an Open Source DAM as Saas platform
Building and Deploying MediaSalsa, an Open Source DAM as Saas platformBuilding and Deploying MediaSalsa, an Open Source DAM as Saas platform
Building and Deploying MediaSalsa, an Open Source DAM as Saas platform
 
ADDO 2019: Looking back at over 10 years of Devops
ADDO 2019:    Looking back at over 10 years of DevopsADDO 2019:    Looking back at over 10 years of Devops
ADDO 2019: Looking back at over 10 years of Devops
 
Can we fix dev-oops ?
Can we fix dev-oops ?Can we fix dev-oops ?
Can we fix dev-oops ?
 
Pipeline as Code
Pipeline as CodePipeline as Code
Pipeline as Code
 
Is there a Future for devops ?
Is there a Future for devops   ? Is there a Future for devops   ?
Is there a Future for devops ?
 
Pipeline as code for your infrastructure as Code
Pipeline as code for your infrastructure as CodePipeline as code for your infrastructure as Code
Pipeline as code for your infrastructure as Code
 
Dev secops opsec, devsec, devops ?
Dev secops opsec, devsec, devops ?Dev secops opsec, devsec, devops ?
Dev secops opsec, devsec, devops ?
 
From MonitoringSucks to Monitoring Love , 2016 Edition
From MonitoringSucks to Monitoring Love , 2016 EditionFrom MonitoringSucks to Monitoring Love , 2016 Edition
From MonitoringSucks to Monitoring Love , 2016 Edition
 
Moby is killing your devops efforts
Moby is killing your devops effortsMoby is killing your devops efforts
Moby is killing your devops efforts
 
Is there a future for devops ?
Is there a future for devops ?Is there a future for devops ?
Is there a future for devops ?
 
Velocity 2011: Production Begins in Development
Velocity 2011: Production Begins in DevelopmentVelocity 2011: Production Begins in Development
Velocity 2011: Production Begins in Development
 
Automating MySQL operations with Puppet
Automating MySQL operations with PuppetAutomating MySQL operations with Puppet
Automating MySQL operations with Puppet
 
Nightmare on Docker street
Nightmare on Docker streetNightmare on Docker street
Nightmare on Docker street
 
From devoops to devops
From devoops to devopsFrom devoops to devops
From devoops to devops
 
Devops is a Security Requirement
Devops is a Security RequirementDevops is a Security Requirement
Devops is a Security Requirement
 
Devops is dead, Long Live Devops
Devops is dead, Long Live DevopsDevops is dead, Long Live Devops
Devops is dead, Long Live Devops
 

Similar a Help , My Datacenter is on fire

Building real time Data Pipeline using Spark Streaming
Building real time Data Pipeline using Spark StreamingBuilding real time Data Pipeline using Spark Streaming
Building real time Data Pipeline using Spark Streamingdatamantra
 
Task migration using CRIU
Task migration using CRIUTask migration using CRIU
Task migration using CRIURohit Jnagal
 
Scaling Monitoring At Databricks From Prometheus to M3
Scaling Monitoring At Databricks From Prometheus to M3Scaling Monitoring At Databricks From Prometheus to M3
Scaling Monitoring At Databricks From Prometheus to M3LibbySchulze
 
How Klout migrated from CDH3 to CDH4 …and survived to tell about it
How Klout migrated from CDH3 to CDH4 …and survived to tell about itHow Klout migrated from CDH3 to CDH4 …and survived to tell about it
How Klout migrated from CDH3 to CDH4 …and survived to tell about itIan Kallen
 
Server fleet management using Camunda by Akhil Ahuja
Server fleet management using Camunda by Akhil AhujaServer fleet management using Camunda by Akhil Ahuja
Server fleet management using Camunda by Akhil Ahujacamunda services GmbH
 
Webinar: Building a multi-cloud Kubernetes storage on GitLab
Webinar: Building a multi-cloud Kubernetes storage on GitLabWebinar: Building a multi-cloud Kubernetes storage on GitLab
Webinar: Building a multi-cloud Kubernetes storage on GitLabMayaData Inc
 
THE RISE AND FALL OF SERVERLESS COSTS - TAMING THE (SERVERLESS) BEAST
THE RISE AND FALL OF SERVERLESS COSTS - TAMING THE (SERVERLESS) BEASTTHE RISE AND FALL OF SERVERLESS COSTS - TAMING THE (SERVERLESS) BEAST
THE RISE AND FALL OF SERVERLESS COSTS - TAMING THE (SERVERLESS) BEASTOpher Dubrovsky
 
Apache Cassandra at Target - Cassandra Summit 2014
Apache Cassandra at Target - Cassandra Summit 2014Apache Cassandra at Target - Cassandra Summit 2014
Apache Cassandra at Target - Cassandra Summit 2014Dan Cundiff
 
Webinar: Choosing the Right Shard Key for High Performance and Scale
Webinar: Choosing the Right Shard Key for High Performance and ScaleWebinar: Choosing the Right Shard Key for High Performance and Scale
Webinar: Choosing the Right Shard Key for High Performance and ScaleMongoDB
 
Piano Media - approach to data gathering and processing
Piano Media - approach to data gathering and processingPiano Media - approach to data gathering and processing
Piano Media - approach to data gathering and processingMartinStrycek
 
AWS Techniques and lessons writing low cost autoscaling GitLab runners
AWS Techniques and lessons writing low cost autoscaling GitLab runnersAWS Techniques and lessons writing low cost autoscaling GitLab runners
AWS Techniques and lessons writing low cost autoscaling GitLab runnersAnthony Scata
 
Cassandra Day London 2015: The Resilience of Apache Cassandra
Cassandra Day London 2015: The Resilience of Apache CassandraCassandra Day London 2015: The Resilience of Apache Cassandra
Cassandra Day London 2015: The Resilience of Apache CassandraDataStax Academy
 
Kubernetes Forum Seoul 2019: Re-architecting Data Platform with Kubernetes
Kubernetes Forum Seoul 2019: Re-architecting Data Platform with KubernetesKubernetes Forum Seoul 2019: Re-architecting Data Platform with Kubernetes
Kubernetes Forum Seoul 2019: Re-architecting Data Platform with KubernetesSeungYong Oh
 
What makes me to migrate entire VPC JAWS PANKRATION 2021
What makes me to migrate entire VPC JAWS PANKRATION 2021What makes me to migrate entire VPC JAWS PANKRATION 2021
What makes me to migrate entire VPC JAWS PANKRATION 2021Naomi Yamasaki
 
Google Cloud - Stand Out Features
Google Cloud - Stand Out FeaturesGoogle Cloud - Stand Out Features
Google Cloud - Stand Out FeaturesGDG Cloud Bengaluru
 
Tapjoy OpenStack Summit Paris Breakout Session
Tapjoy OpenStack Summit Paris Breakout SessionTapjoy OpenStack Summit Paris Breakout Session
Tapjoy OpenStack Summit Paris Breakout SessionWeston Jossey
 
NetflixOSS Meetup season 3 episode 1
NetflixOSS Meetup season 3 episode 1NetflixOSS Meetup season 3 episode 1
NetflixOSS Meetup season 3 episode 1Ruslan Meshenberg
 
Your Testing Is Flawed: Introducing A New Open Source Tool For Accurate Kuber...
Your Testing Is Flawed: Introducing A New Open Source Tool For Accurate Kuber...Your Testing Is Flawed: Introducing A New Open Source Tool For Accurate Kuber...
Your Testing Is Flawed: Introducing A New Open Source Tool For Accurate Kuber...StormForge .io
 
Processing TeraBytes of data every day and sleeping at night
Processing TeraBytes of data every day and sleeping at nightProcessing TeraBytes of data every day and sleeping at night
Processing TeraBytes of data every day and sleeping at nightLuciano Mammino
 

Similar a Help , My Datacenter is on fire (20)

Building real time Data Pipeline using Spark Streaming
Building real time Data Pipeline using Spark StreamingBuilding real time Data Pipeline using Spark Streaming
Building real time Data Pipeline using Spark Streaming
 
Task migration using CRIU
Task migration using CRIUTask migration using CRIU
Task migration using CRIU
 
Scaling Monitoring At Databricks From Prometheus to M3
Scaling Monitoring At Databricks From Prometheus to M3Scaling Monitoring At Databricks From Prometheus to M3
Scaling Monitoring At Databricks From Prometheus to M3
 
How Klout migrated from CDH3 to CDH4 …and survived to tell about it
How Klout migrated from CDH3 to CDH4 …and survived to tell about itHow Klout migrated from CDH3 to CDH4 …and survived to tell about it
How Klout migrated from CDH3 to CDH4 …and survived to tell about it
 
Server fleet management using Camunda by Akhil Ahuja
Server fleet management using Camunda by Akhil AhujaServer fleet management using Camunda by Akhil Ahuja
Server fleet management using Camunda by Akhil Ahuja
 
Cloud arch patterns
Cloud arch patternsCloud arch patterns
Cloud arch patterns
 
Webinar: Building a multi-cloud Kubernetes storage on GitLab
Webinar: Building a multi-cloud Kubernetes storage on GitLabWebinar: Building a multi-cloud Kubernetes storage on GitLab
Webinar: Building a multi-cloud Kubernetes storage on GitLab
 
THE RISE AND FALL OF SERVERLESS COSTS - TAMING THE (SERVERLESS) BEAST
THE RISE AND FALL OF SERVERLESS COSTS - TAMING THE (SERVERLESS) BEASTTHE RISE AND FALL OF SERVERLESS COSTS - TAMING THE (SERVERLESS) BEAST
THE RISE AND FALL OF SERVERLESS COSTS - TAMING THE (SERVERLESS) BEAST
 
Apache Cassandra at Target - Cassandra Summit 2014
Apache Cassandra at Target - Cassandra Summit 2014Apache Cassandra at Target - Cassandra Summit 2014
Apache Cassandra at Target - Cassandra Summit 2014
 
Webinar: Choosing the Right Shard Key for High Performance and Scale
Webinar: Choosing the Right Shard Key for High Performance and ScaleWebinar: Choosing the Right Shard Key for High Performance and Scale
Webinar: Choosing the Right Shard Key for High Performance and Scale
 
Piano Media - approach to data gathering and processing
Piano Media - approach to data gathering and processingPiano Media - approach to data gathering and processing
Piano Media - approach to data gathering and processing
 
AWS Techniques and lessons writing low cost autoscaling GitLab runners
AWS Techniques and lessons writing low cost autoscaling GitLab runnersAWS Techniques and lessons writing low cost autoscaling GitLab runners
AWS Techniques and lessons writing low cost autoscaling GitLab runners
 
Cassandra Day London 2015: The Resilience of Apache Cassandra
Cassandra Day London 2015: The Resilience of Apache CassandraCassandra Day London 2015: The Resilience of Apache Cassandra
Cassandra Day London 2015: The Resilience of Apache Cassandra
 
Kubernetes Forum Seoul 2019: Re-architecting Data Platform with Kubernetes
Kubernetes Forum Seoul 2019: Re-architecting Data Platform with KubernetesKubernetes Forum Seoul 2019: Re-architecting Data Platform with Kubernetes
Kubernetes Forum Seoul 2019: Re-architecting Data Platform with Kubernetes
 
What makes me to migrate entire VPC JAWS PANKRATION 2021
What makes me to migrate entire VPC JAWS PANKRATION 2021What makes me to migrate entire VPC JAWS PANKRATION 2021
What makes me to migrate entire VPC JAWS PANKRATION 2021
 
Google Cloud - Stand Out Features
Google Cloud - Stand Out FeaturesGoogle Cloud - Stand Out Features
Google Cloud - Stand Out Features
 
Tapjoy OpenStack Summit Paris Breakout Session
Tapjoy OpenStack Summit Paris Breakout SessionTapjoy OpenStack Summit Paris Breakout Session
Tapjoy OpenStack Summit Paris Breakout Session
 
NetflixOSS Meetup season 3 episode 1
NetflixOSS Meetup season 3 episode 1NetflixOSS Meetup season 3 episode 1
NetflixOSS Meetup season 3 episode 1
 
Your Testing Is Flawed: Introducing A New Open Source Tool For Accurate Kuber...
Your Testing Is Flawed: Introducing A New Open Source Tool For Accurate Kuber...Your Testing Is Flawed: Introducing A New Open Source Tool For Accurate Kuber...
Your Testing Is Flawed: Introducing A New Open Source Tool For Accurate Kuber...
 
Processing TeraBytes of data every day and sleeping at night
Processing TeraBytes of data every day and sleeping at nightProcessing TeraBytes of data every day and sleeping at night
Processing TeraBytes of data every day and sleeping at night
 

Más de Kris Buytaert

Years of (not) learning , from devops to devoops
Years of (not) learning , from devops to devoopsYears of (not) learning , from devops to devoops
Years of (not) learning , from devops to devoopsKris Buytaert
 
Observability will not fix your Broken Monitoring ,Ignite
Observability will not fix your Broken Monitoring ,IgniteObservability will not fix your Broken Monitoring ,Ignite
Observability will not fix your Broken Monitoring ,IgniteKris Buytaert
 
Infrastructure as Code Patterns
Infrastructure as Code PatternsInfrastructure as Code Patterns
Infrastructure as Code PatternsKris Buytaert
 
From devoops to devops 13 years of (not) learning
From devoops to devops 13 years of (not) learningFrom devoops to devops 13 years of (not) learning
From devoops to devops 13 years of (not) learningKris Buytaert
 
10 Years of #devopsdays weirdness
10 Years of #devopsdays weirdness10 Years of #devopsdays weirdness
10 Years of #devopsdays weirdnessKris Buytaert
 
Continuous Infrastructure First Ignite Edition
Continuous Infrastructure First  Ignite EditionContinuous Infrastructure First  Ignite Edition
Continuous Infrastructure First Ignite EditionKris Buytaert
 
Looking back at 5 years of #cfgmgmtcamp
Looking back at 5 years of #cfgmgmtcampLooking back at 5 years of #cfgmgmtcamp
Looking back at 5 years of #cfgmgmtcampKris Buytaert
 
The Return of the Dull Stack Engineer
The Return of the Dull Stack EngineerThe Return of the Dull Stack Engineer
The Return of the Dull Stack EngineerKris Buytaert
 
Looking back at 7.5 years of Devopsdays , DOd PDX
Looking back at 7.5 years of Devopsdays , DOd PDXLooking back at 7.5 years of Devopsdays , DOd PDX
Looking back at 7.5 years of Devopsdays , DOd PDXKris Buytaert
 
Devopsdays Amsterdam 2017 Keynote, looking back at 5 years of AMS
Devopsdays Amsterdam 2017 Keynote, looking back at 5 years of AMSDevopsdays Amsterdam 2017 Keynote, looking back at 5 years of AMS
Devopsdays Amsterdam 2017 Keynote, looking back at 5 years of AMSKris Buytaert
 

Más de Kris Buytaert (10)

Years of (not) learning , from devops to devoops
Years of (not) learning , from devops to devoopsYears of (not) learning , from devops to devoops
Years of (not) learning , from devops to devoops
 
Observability will not fix your Broken Monitoring ,Ignite
Observability will not fix your Broken Monitoring ,IgniteObservability will not fix your Broken Monitoring ,Ignite
Observability will not fix your Broken Monitoring ,Ignite
 
Infrastructure as Code Patterns
Infrastructure as Code PatternsInfrastructure as Code Patterns
Infrastructure as Code Patterns
 
From devoops to devops 13 years of (not) learning
From devoops to devops 13 years of (not) learningFrom devoops to devops 13 years of (not) learning
From devoops to devops 13 years of (not) learning
 
10 Years of #devopsdays weirdness
10 Years of #devopsdays weirdness10 Years of #devopsdays weirdness
10 Years of #devopsdays weirdness
 
Continuous Infrastructure First Ignite Edition
Continuous Infrastructure First  Ignite EditionContinuous Infrastructure First  Ignite Edition
Continuous Infrastructure First Ignite Edition
 
Looking back at 5 years of #cfgmgmtcamp
Looking back at 5 years of #cfgmgmtcampLooking back at 5 years of #cfgmgmtcamp
Looking back at 5 years of #cfgmgmtcamp
 
The Return of the Dull Stack Engineer
The Return of the Dull Stack EngineerThe Return of the Dull Stack Engineer
The Return of the Dull Stack Engineer
 
Looking back at 7.5 years of Devopsdays , DOd PDX
Looking back at 7.5 years of Devopsdays , DOd PDXLooking back at 7.5 years of Devopsdays , DOd PDX
Looking back at 7.5 years of Devopsdays , DOd PDX
 
Devopsdays Amsterdam 2017 Keynote, looking back at 5 years of AMS
Devopsdays Amsterdam 2017 Keynote, looking back at 5 years of AMSDevopsdays Amsterdam 2017 Keynote, looking back at 5 years of AMS
Devopsdays Amsterdam 2017 Keynote, looking back at 5 years of AMS
 

Último

Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024The Digital Insurer
 
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot TakeoffStrategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoffsammart93
 
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ..."I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...Zilliz
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FMESafe Software
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...apidays
 
ICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesrafiqahmad00786416
 
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...Zilliz
 
Exploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with MilvusExploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with MilvusZilliz
 
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...Angeliki Cooney
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingEdi Saputra
 
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...Jeffrey Haguewood
 
Ransomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdfRansomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdfOverkill Security
 
AXA XL - Insurer Innovation Award Americas 2024
AXA XL - Insurer Innovation Award Americas 2024AXA XL - Insurer Innovation Award Americas 2024
AXA XL - Insurer Innovation Award Americas 2024The Digital Insurer
 
presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century educationjfdjdjcjdnsjd
 
CNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In PakistanCNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In Pakistandanishmna97
 
DEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 AmsterdamDEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 AmsterdamUiPathCommunity
 
[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdf[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdfSandro Moreira
 
2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...Martijn de Jong
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...apidays
 
Manulife - Insurer Transformation Award 2024
Manulife - Insurer Transformation Award 2024Manulife - Insurer Transformation Award 2024
Manulife - Insurer Transformation Award 2024The Digital Insurer
 

Último (20)

Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024
 
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot TakeoffStrategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
 
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ..."I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...
 
ICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesICT role in 21st century education and its challenges
ICT role in 21st century education and its challenges
 
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
 
Exploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with MilvusExploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with Milvus
 
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
 
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
 
Ransomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdfRansomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdf
 
AXA XL - Insurer Innovation Award Americas 2024
AXA XL - Insurer Innovation Award Americas 2024AXA XL - Insurer Innovation Award Americas 2024
AXA XL - Insurer Innovation Award Americas 2024
 
presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century education
 
CNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In PakistanCNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In Pakistan
 
DEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 AmsterdamDEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
 
[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdf[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdf
 
2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
 
Manulife - Insurer Transformation Award 2024
Manulife - Insurer Transformation Award 2024Manulife - Insurer Transformation Award 2024
Manulife - Insurer Transformation Award 2024
 

Help , My Datacenter is on fire

  • 1. The great fire of 2021 Kris Buytaert @krisbuytaert
  • 2. @krisbuytaert March 10 , 2021 ● 03:59 Incoming Phone call ● “Our Datacenter is On fire” ● 5 minutes of waking up.. going downstairs trying to realize what was just said to me, partial panic , partial .. it can’t be THAT bad
  • 4. @krisbuytaert Kris Buytaert ● I used to be a Dev, ● Then Became an Op ● CTO and Open Source Consultant @inuits.eu ● Everything is a freaking DNS Problem ● Evangelizing devops ● Organiser of #devopsdays, #cfgmgmtcamp, #loadays, ….
  • 5. @krisbuytaert Impact : ● 3.6 million websites across 464.000 distinct domains ● https://news.netcraft.com/archives/2021/03/10/ovh-fire.html
  • 6. @krisbuytaert Immediate Assesment ? ● What Customer Facing Production services have we lost ? ● What Internal Production services have we lost ? – Do we need any of those to recover ? ● What are the priorities ? – Which services ● 24/7 money generating platforms ● Office hours tooling – What parts of these services ● Basic Service, no redundancy
  • 7. @krisbuytaert Our Immediate visible Impact ● Multiple customers failed over automatically to the 2ndary DC ● 1 VIP address failed to move ● 1 platform was spread over SBG2 & SBG1 ● 1 DR site destroyed
  • 8. @krisbuytaert Failing VIP ● 1 VIP address failed to move (API call failed) ● Manual action to use a different public IP for the loadbalancer ● Updated DNS records ● Except for the 1 domain the customer controlled. ● This cost us about 60 minutes before we could reach someone with the right credentials to update the dns ● Some customer devices had the VIP address hardcoded :( – Manual interventions needed :(
  • 9. @krisbuytaert Customer Impact ● 1 platform was spread over SBG2 & SBG1 ● Rebootstrapped the platform on a different ISP, restored the backup ● At 0900 am platform was ready for use, ● We were improving performance of the platform ● Customer wouldn’t have known if we didn’t tell them
  • 10. @krisbuytaert Total Loss: ● We lost 13 physical servers ● 135 vm’s
  • 11. @krisbuytaert Root Cause ● Klaba does not say the UPS is the definite cause. "We don’t have all the answers today," he said. The OVHcloud staff responded to alarms at 11.42 pm on Tuesday, but the affected part of the data center had already filled with smoke: "Two minutes after, they took the decision to leave, because it was too dangerous." ● The firefighter's thermal cameras found UPS7 and UPS8 on fire, but further data will be extracted from on-site cameras: "We have 300 cameras in Strasbourg," said Klaba. "We expect to have all the answers about how it started. We will give you all the information." ● https://www.datacenterdynamics.com/en/news/ovh-fire-octave-klaba-says-ups-systems- were-ablaze/
  • 12. @krisbuytaert Our Impact NextCloud ● Gitlab ● Lots of development environments ● Lots of Test nodes ● Partial K8s clusters
  • 13. @krisbuytaert We survived ● 18 hours of work ● 2 engineers at night ● 4-5 during daytime ● Recovery ● Communication ● Restoring Backups
  • 14. @krisbuytaert At 0910 ● Another customer seemed to have issues accessing some files ● We had overlooked 1 gluster cluster that seemed healthy but wasn’t ● Switched to a single node mount point (don’t try this at home)
  • 16. @krisbuytaert Phase II ● After 0900 we started restoring our internal tools from backup ● Verified all the secondary nodes would be making backups ● Verified we had sufficient diskspace left ● “Shoemaker always wears the worst shoes” ● Challenge : Hardware !
  • 17. @krisbuytaert Hardware Landrush ● Everybody needed new hardware ● We had spare hardware ready for a new platform in the wrong DC of a different supplier . Temporary OK ● We spun up new hardware at a different supplier
  • 18. @krisbuytaert Wrong decisions ● On day 2 we wasted at least 6 hours trying spinning up a new supplier ● Figuring out their network and redundancy strategies was not the right time ● We could scale on the existing suppliers
  • 19. @krisbuytaert Phase III ● Restoring the rest of the services ● What boxen would we be getting back ? ● What boxen were foobar ? ● (in the end we only got 3 physical servers back which were in SBG3), the rest was lost ● Priority on making sure all pipeline promotions of actively developed platforms were running again ● Hardware availability
  • 20. @krisbuytaert The day after ● New DR platform where needed ● Inventory of what besides production was impacted (lots of development boxen, partial clusters) ● Prioritization of what needs to be 100% back first ● What hardware do we have / need additionally
  • 21. @krisbuytaert Your NEW DR Plan ● You have just lost your DR site ● You need to plan again – Do we take the risk for now ? – Do we respin a new DR ? ● IAC + Datasync makes DR Trivial
  • 22. @krisbuytaert What saved us ? ● Real Infrastructure as Code ● Architecture ● Backups ● Fast escalation
  • 23. @krisbuytaert Real Infrastructure as Code ● Desired State => Puppetize all the things ● 100% control , no 3rd parties involved – Provision vm – Deploy applications – Deploy database schemas ● No handovers , 1 person can deploy this, no other teams involved ● Exported resources for Loadbalancers (haproxy), Monitoring (icinga), Databases (mysql)etc – Heavy automation ● All deployed versions are available in (yum) repositories ● Everything is a Pipeline Puppet, Hiera, DNS, ... (jenkins)
  • 24. @krisbuytaert CloudNative vs Cloudnaive ● We don’t own hardware ● Baremetal on demand at $ < AWS – Hetzner / ovh / ... ● Spin up 120 seconds, decomission when unneeded. ● vm definitions are in code ● Ansible to bootstrap, Puppet for Desired state
  • 25. @krisbuytaert Multi Cloud ● OVH ● Hetzner ● ... ● Workload is spread ● Customer DR is in other Supplier
  • 26. @krisbuytaert Multi Datacenter ● OVH – SBG, GRA, RBX, .. ● Hetzner – FSN X, HEL
  • 27. @krisbuytaert CloudNative ? ● Critical Customer services are build redundantly, (corosync, haproxy, mysql, elastic) ● Losing a bare metal should doesnot have an impact ● We can rebootstrap our nodes (kickstart+ puppet + ansible) ● All UGC data we know about is backed up (rsnapshot)
  • 28. @krisbuytaert Cloud Agnostic ● A vm is a vm is a vm ● Vendor Specific Features – e.g VIP addresses are abstracted in code
  • 29. @krisbuytaert Dedicated stacks per Customer/Project ● Single Purpose vm’s ● Single Purpose Database Clusters ● Single Purpose Storage Clusters ● Single Purpose Loadbalancers => Reduced impact, Reduced complexity
  • 30. @krisbuytaert Clusters are Multi DC ● Depending on scale , preference on HA vs actual Loadbalancing ● Pinning of resources within 1 DC ● Lessons Learned: we need Multi Campus
  • 31. @krisbuytaert HA Applications ● No local file usage – If local files -> {gluster,drbd,...} ● MySQL replication ● Elastic multiple nodes (if we can’t generate the data) ● Multiple instances of rabbitmq
  • 32. @krisbuytaert HA Storage ● VM’s on Local Disk – 100% puppetized => no differences in config/ software deployment – “Disposable” ● Storage distributed – Gluster, DRBD, (Ceph) ● (No Raid)
  • 33. @krisbuytaert Backups ● Are both on site ● And off Site
  • 35. @krisbuytaert Lessons Learned ● Make sure you can always provision more resources on multiple providers ● A Datacenter is not always Datacenter, even with different buildings you need Campus redundancy ● Fix the shoemaker problem ● Everything IS a fscking dns problem – U WANT control over the DNS of the production sites you run ● Or an MTU problem
  • 36. @krisbuytaert Thnx Thnx to everyone who helped out during the outage !
  • 38. @krisbuytaert Contact Inuits Inuits Essensteenweg 31 Essensteenweg 31 Brasschaat Brasschaat Belgium Belgium 891.514.231 891.514.231 +32 475 961221 +32 475 961221 Kris Buytaert Kris.Buytaert@inuits.eu Kris Buytaert Kris.Buytaert@inuits.eu Further Reading Further Reading @krisbuytaert @krisbuytaert http://www.krisbuytaert.be/blog/ http://www.krisbuytaert.be/blog/ https://inuits.eu/ https://inuits.eu/