SQL Server High Availability and DR - Too Many Choices!
TOO MANY CHOICES!!
• SQL Server MVP for 8 years
• Consultant + Founder (Straight Path Solutions)
• 20+ Years working with SQL Server
• Blog at www.straightpathsql.com
• Notes from today’s session is the current blog post.
• E-mail me anytime: email@example.com
Our Journey Today: The What
• This is an Overview
• We’ll talk about the various terms (HA/DR/RPO/RTO/SLA)
• We’ll talk about the approaches available for HA/DR
• We’ll talk a bit about the cloud
• We’ll get into the pros and cons of the approaches
• We’ll interact, I hope!
• You’ll leave with a sense of direction
• All Discussion/Interaction/PowerPoint
Our Journey Today: The How/When
• 20-30 Minutes of Slides
• 20-30 Minutes of Q&A – ask me your questions. I’ll answer
questions (hopefully!!) and take notes on the Q&A.
• Leave. Go enjoy the raffles!
• Blog Post: I’ll write up a blog post from the Q&A, include the
slides, and the discussion points so you have a reference.
Three Real Clients
• FCI + AG + Lack of Best Practices = KABOOM! –
on Tax Day!!
• Ransomware… Highly Available Ransomware….
• 1,200 DBs heading to AWS. Only choice is lots of
AGs, right? (no SAN, max DBs in an AG?)
• Recovery Point Objective (RPO) – To what point can you recover?
• Recovery Time Objective (RTO) – How long will you be down?
• Service Level Agreement (SLA) – What did you promise?
• SQL Server Instance – A single installation of SQL Server – what
you connect to in SSMS. Your jobs, alerts, linked servers, logins,
etc all live at the Instance level.
• SQL Server Database – A single database on that instance.
High Availability <> Disaster Recovery
• Oh No!!
• Server dies
• Need to be up really fast
• Quite often folks want this
• Oh $!%!@$ !!!
• Data Center (or SAN or core
network to data center/etc
• Typically, can have a bit more
time (can still be quite fast) to
• Can be manual
You Have to Talk… To The Business…
Never let the business decide alone on your HA/DR
solution, but never decide without them.
• How long can we be down for?
• How much data can we lose?
• (Those two questions are a battle of wits, for sure! Well wits and
dollars. Fight it with the dollars angle)
• What happens if we lose the primary data center?
• What about everything else?
• What about planned maintenance and patching? Is there a
window for that?
• Do we care about “C drive” protection?
Technologies Available – The Many
• In SQL:
• Failover Cluster Instances
• Availability Groups
• Log Shipping
• (Replication – but….)
• Outside of SQL:
• Fault Tolerant Hardware (but….)
• VMWare/HyperV HA options
• VMWare/HyperV DR options
• Storage Replication
• Tools like Azure Site Recovery
• Platform as a Service Cloud offerings (RDS/Azure SQL/Azure SQL MI)
Some Common Approaches
Method HA DR Passive
Easy Failback L
Log Shipping N Y Sort Of N $$
Always On Y Y Y Y $$$ (can now do
Standard, but let’s
Y With SIOS
Replica it can
Yes Very $$$
VMWare HA Y N (w/ third party
offerings you can)
No It Depends $$$
Third Party Tools
N Y Sort Of Usually $$-$$$
N Maybe – if your
RTO supports it
No No $
Instance Level Protection vs DB Level
Technologies which protect an entire instance (or server) protect
everything on your instance. Your jobs, logins, all DBs (even new
ones, out of the box), all protected. DB Level protection means you
are only protecting the database(s)
Failover Cluster INSTANCE
• Instance Level Protection
• Automatic Failover (really the only option out of the box)
• Needs to either have shared storage (traditionally this was a hard and
fast requirement) or the appearance of shared storage.
• Typically was always in one data center, HA only, but now can stretch
data centers, be SAN-less, have different IPs and even domainless.
• I’m starting to lean here more these days again
• SQL Server Standard – two node Single instance
• SQL Server Enterprise – multi node – multi instance
• Failover seamless to users
• With SQL 2016+ now available in Standard (*****)
• Database Level Protection (you are responsible for the logins,
jobs, alerts, linked server, new DBs, etc – DBATools.IO is your
• Automatic Failover available for synchronous Ags
• Can do asynchronous to cross data centers
• Can be in different subnets
• Can now stretch domains or even be domainless
• AG Listener and seamless to users failover
• Read Only workload and automatic routing of it
• Seriously – a tried, trusted, and good DR option
• Database level
• Some manual work (or scriptable yourself lots of help on the web
there for it)
• It’s just log backup, copy, restore with norecovery or standby
• Can sort of kind of maybe read off of the secondary
• Great for migrations as well
Speaking of Licenses
• Disclaimer: I make no warranties in license conversations. The
Microsoft licensing guide is a lot lake the iocane powder scene in
the Princess Bride… Slightly less complicated than the tax code.
• Enterprise vs Standard
• Software Assurance for a “Free failover”
• “Free failover” covers ONE Node.
• If you are reading from it? You are licensing it.
• Enterprise is $$$$
Outside of SQL Server
• HA/DR At the Virtualization Layer
• In the Cloud
• Combing technologies
A Word on Backup and Recovery
• Lots of great sessions here about this topic already.
• The Short Story:
• These technologies do not eliminate your need to backup.
• These technologies do not eliminate your need to test your backups.
• These technologies do not eliminate your need for Backups to be inaccessible
to bad people!
• These technologies do not eliminate your need to do Integrity Checks.
• Backup is still your responsibility.
• A Backup is still useless if you can’t restore it and haven’t tested,
practiced and understood the restore process.
• HA/DR will take a rogue update and make that rogue update highly
available and fault tolerant!
Now. Now You Talk to the Business
• Take the questions and approaches above
• Take the licensing information
• Take your existing infrastructure and capabilities
• Come up with a chart and give them options
• Show them RPO/RTO comparisons
• Show them Pricing comparisons
• Maybe 3 tiers – Gold/Silver/Bronze
• Maybe you have one of each to consolidate towards?
Q&A / Interaction
• I’ll write down notes from our conversation and have a blog post up by end of
Tuesday with updates based on your questions. Straightpathsql.com/blog -
I’ll tweet about it also – @mike_walsh.
• What is on your mind?
• Let’s get into challenges/questions/etc.
• mike@StraightPathSQL.com – open door policy.
• Blog post from Q&A – www.StraightPathSQL.com/blog