webinar presentation: stories of accidental data loss
TRANSCRIPT
Confidential and Proprietary1
Did I Do That? Stories of Accidental Data Loss
Confidential and Proprietary2
Agenda Real-world Data Loss Incidents
Lessons Learned
Business Impacts of Inadequate Data Protection
Case Studies on Data Protection
Architecture Considerations for Backup & Recovery
Confidential and Proprietary3
Incident #1 – Multiple Backup Solutions Do Not Ensure Protection
Production – db1Issues, Merge Requests, Code Snippets
Standby– db2
• 6 Hours of Data Lost
• ~40 Hours of Downtime
Data Protection Production Issue Restore Issue• Snapshots• NFS Backups• S3 Backups• Database Dump• Replication
• Spammers destabilized DB• Replication stopped
working• DBA deleted files on
production DB instead of standby
• Replication not working• Snapshot 24 hours old• NFS backup could not be
located• Database dump failing• Backup to S3 not working
Confidential and Proprietary4
Incident #2 – Multiple Replicas Do Not Ensure Protection
• 2 weeks of downtime• Resources working
24x7• Total Cost - $1.1M Hadoop Cluster
Multi-PetabyteHIVE
Data
Inge
st
Data Protection Production Issue Restore Issue
• Relying on 3 replicas in Hadoop
• Developer cleaning old tables• Accidentally dropped active
table• 500 TB data lost
• All replicas deleted• No backups• Rebuilt table by re-ingesting
all data
Confidential and Proprietary5
Incident #3 – Immutable Storage Systems Do Not Ensure Protection
• Regulatory certification denied• Large fines
CRM Application
Secure Storage for Contracts
Retention Management
Data Protection Production Issue Restore Issue• Secure storage
environment with retention management
• Sales rep changed retention for the wrong customer from 24 mos. to 6 mos.
• 18 months of contracts deleted from storage
• No backups resulting in permanent data loss
Confidential and Proprietary6
Lessons Learned
Production Environment
Use descriptive database names
No testing in production databases (limited access)
Use multiple replicas to protect from hardware failure – NOT BACKUP
Use remote replication for Disaster Recovery – NOT BACKUP
Backup Architecture Design
Not all data is equalDesign based on Recovery-Time-Objectives & Recovery-Point-Objectives (RPO)Store backups off-hostKeep multiple restore pointsKnow the cost of recreating lost data
Operations
Create and maintain a playbook
Monitor backup processes
Regularly practice recovery scenarios
Confidential and Proprietary7
Business Impact of Inadequate Data Protection
Extended Application Downtime• Revenue loss• Productivity loss• Lose customers to competitors• Damage to reputation/brandData Loss• Valuable resources spent on rebuilding data• Regulatory & compliance violations resulting in
fines• Customer frustration• Lawsuits
Confidential and Proprietary8
Data Protection Case Study #1
Amazon EC2Cassandra Production Cluster
52 nodes, 58 Terabytes
BACKUP
Backup/Recovery Challenges• Backing up and restoring a 58 TB
table• Updating QA/DR cluster with
production data
Solution Characteristic• Combined backup & DR• Incremental-forever backups & restores• Storage de-duplication & compression• Backup to Amazon S3
Amazon EC2Cassandra DR Cluster
36 nodes, 58 Terabytes
DR
Confidential and Proprietary9
Talena GUI
Five node Talena
cluster usingdirect-attached
storage
Vertica Cluster10-node
Vertica Cluster10-node
HadoopCluster
Data Protection Case Study #2
Backup/Recovery Challenges
• Native backup tool had limitations
• Multiple scripts hard to manage
Solution Characteristics• Hadoop Namenode backup• Topology-agnostic Vertica
backup/recovery• Single pane of glass
Confidential and Proprietary10
Data Protection Case Study #3
Hadoop Cluster37 nodes, 425
TB
Hadoop Cluster8-node, 50 TB
Hadoop Cluster16 nodes, 180
TB
PROTECT
ARCHIVE
BACKUP
ARCHIVEBACKUP
BACKUP
Backup/Recovery Challenges• Complexity of managing
600+TB of backups• Unable to meet RPO
requirements
Solution Characteristics• Single pane of glass for all Hadoop
clusters• Hourly backup to meet stringent RPO
needs• Scale-out solution to accommodate
scale
Talena GUI
Confidential and Proprietary11
Architectural Considerations for Backup & Recovery
Dealing with Distributed Deployments• Scale-out solution• Parallel data transfers between primary & backup
Storage• An agentless modelDealing with Volume & Velocity• De-duplicated, compressed backups• Scalable, searchable catalog that can store millions of
objects• Incremental-forever backups with single-step restore• Fast & granular recoveryDealing with Data Variety• Application-aware backups and restores• Content-aware storage optimization
Confidential and Proprietary12
Q&AWe’ll send you a link to our eBook “Compendium of Data Loss Horror Stories”
Additional resources: talena-inc.com/resources and talena-inc.com/blog
Ping us with any additional questions: [email protected]
Confidential and Proprietary13
THANK YOU