Backup Strategies for Platinum

1
Backup and Restore Strategies for PlatinumJerry
CochranManager Exchange Solutions
EngineeringCompaq Computer Corporation
2
(No Transcript)
3
Agenda

• Exchange Technology Foundation
• Current Issues
• Exchange 2000 Storage
• Exchange 2000 Backup/Restore
• Exchange 2000 DR Demo
• Best Practices

4
Exchange Server Basics

• JET/ESE Technology
• Transaction-based Storage
• Database Logs
• Online Backup/Restore API
• Integrity Checking of the Database

5
Current Disaster Recovery Issues

• Large Information Stores
• Disaster Recovery Limits Scalability
• Backup and Restore Time
• Database Corruption Potential
• Lack of Storage Management Flexibility

6
Exchange 2000 Storage

• New Administration and Deployment Paradigms
• Control, Complexity, Flexibility
• New Concepts
• Storage Group (SG)
• Multiple SGs and MDBs
• New Streaming File
• MDBEDBSTM
• Multiple Public Folder Hierarchies

7
Exchange 2000 Storage

• Reliability, Security, Manageability
• Partition or Segment User Data
• Multiple companies on same server
• Separate SG or Database
• Special Mailboxes
• CEO, VPs, etc.
• Large Public Folder applications
• KM and Portal Applications

8
Storage Group

• Each SG runs as a JET/ESE instance within
  STORE.EXE
• Each SG has its own set of Tx logs
• Unit of Failover for Cluster
• Limitations per server
• 15 SGs
• 6 MDBs per ISG
• 1 SG and 1 MDB reserved
• Recovery
• Separate from limits above

9
Property Database (EDB)

• Properties Database
• Contains Folder Tables for messaging data
  (messages)
• MAPI Messages Attachments
• Similar to Exchange 5.5 (ESE97)
• Headers for STM pages

10
Streaming Database (STM)

• STM file introduced to hold native Internet
  content
• Suitable for voice, video, etc.
• Automatic format conversion when accessed by MAPI
  clients
• NOT USED BY MAPI CLIENTS
• Clustered block (4K) run format
• 64K runs (4K times 16)
• Headers kept in EDB file (i.e. checksum)
• Depends on Format Type (i.e. MIME)

11
Transaction Log (.log) Files

• What is a current database

.edb file .stm file (Database)
CURRENT
Uncommited Entries in Transaction Logs (.Log)
12
Patch (.pat) Files

• Patch (.pat) Files Are Only Used During Online
  Backup
• Occur when page updates cause splits

Committed
.pat File
Uncommitted
.edb File
13
Checkpoint (.CHK) Files

• Track Location of last committed transaction
• Needed for efficient recovery

Committed
Checkpoint (EDB.CHK)
Uncommitted
Transaction Log File
14
Recovery Scenarios

• Complete Server Restore
• Windows 2000
• IIS Metabase
• Exchange 2000
• Information Stores
• Information Store Recovery
• Storage Group
• Database (MDB)
• Mailbox
• Message

15
Exchange 2000 Backup- Whats new

• New Stuff to Backup
• SG, STM, IIS, SRS, KMS
• Enhanced Backup API
• New functionality
• Concurrent Operations
• Parallel Backup/Restore on SGs
• Granularity
• SG or MDB

16
Online Backup Types
Type Data Logs Restore Backed Up Purged
Normal (full) Incremental Differential Snapshot
Copy
Database log files Log files only Log files
only Implementation Database
Yes Yes No No No

• Last Normal
• Start Service
• Last Normal
• Every Incr
• Start Service
• Last Normal
• Last Diff
• Start Service
• SnapRestore
• Log Restore
• Soft Recover
• Database PIT

17
Exchange 2000 BackupProcess flow
Backup APIs Called
ESE Backup Mode
Begin Backup
3rd party agent makes API calls to inform
Store of backup procedure Type of backup

• Store informs ESE
• and backup mode entered
• PAT File is created for
• Each database in backup
• New log generation started

Agent requests DB pages in sequence Pages
are check-summed as read Page splits written
to PAT
ESE Normal Mode
End Backup
Backup Complete
Page read completed Logs copied to SET PAT
files copied to SET Logs Truncated SET Closed
18
Strategies - Backup
Sunday
Monday
Tuesday
Wednesday
Thursday
Daily normal backup
Normal with incremental
Normal with differential
19
Exchange 2000 Restore

• Faster
• Partitioned User Data
• Log Replay
• Less User Impact
• Only Users affected
• New ESE Instance
• Manage restore and log replay

20
Exchange 2000 RestoreProcess Flow
Dismount Database or ISG
ESE Restore Mode
Begin Restore
Agent copies EDB/STM from Tape to DB path
Log files from SET are copied To temp restore
location PAT files copied from SET to Temp
location
Store informs ESE and restore mode entered
Restore SG created
3rd party agent or Administrator
ESE Normal Mode
End Backup
Restore Complete
PAT file processed by ESE restore instance
SET Logs are processed By ESE restore
instance Current logs processed By ESE
restore instance Cleanup/Restore SG killed
DB is mounted by SG
21
Strategies - Restore
1 Tape
n Tapes
2 Tapes
..
Daily Normal
Normal With incremental
Normal With differential
Tapes, Time, Volume
22
Snapshots

• API support in Exchange 2000
• Vendor Implementation specific
• Not an Excuse for larger stores

23
Snapshot Backup

• Online access via APIs
• Call Snapshot APIs
• Commit dirty pages to disk
• Take Snapshot of EDB/STM
• Use APIs to copy out log files
• Close backup

24
Snapshot Restore

• Online access via APIs
• Restore Snapshot
• Restore Logs associated with Snapshot to temp
  location
• ESE replay logs to current
• Soft Recovery
• May also restore to snapshot point
• Point in time without current logs
• Vendor Implementation Specific

25
Exchange 2000 Demo

• Backup and Restore

26
Best PracticesPlanning

• Understand Storage Groups and DBs
• Plan Size, Number, Growth, DR impact
• Prepare Disaster Recovery Plan
• Devices, Routine, Rotation, Verification
• Review and Update the Plan
• Exchange 2000 is a good reason
• Disaster Recovery Toolkit
• Configuration Management

27
Best PracticesOperations

• Scenario Training
• Catastrophe, Store, Mailbox, Message
• Recovery Servers
• Part of Production Deployment
• Comprehend Complete System
• Not Just Exchange
• Hardware, OS, Drivers, SPs
• Fire Drills
• Test your Plan!

28
Best PracticesBackup/Restore

• Perform Daily Full/Normal Backup
• Proactively Scan Event logs for errors
• Restore Gotchas
• Selecting Wrong Options
• Example Last Backup Set Option
• Operator Intervention

29
Offline Backup/RestoreJust Say No

• Are Databases Consistent?
• ESEUTIL or ESEFILE
• Managing Logs and Databases
• Thats what the APIs are for!
• Use Extreme Caution
• Wrong place/Wrong time
• Log Replay
• Too much Operator Error Potential
• BEST PRACTICE DONT DO IT!

30
(No Transcript)