VMware Backup & Recovery Doug Hazelman Senior Director, Product Strategy Veeam Software Doug.Hazelman@veeam.com Twitter: @VMDoug
Agenda VMware image level backup overview Considerations for Exchange Backup & Recovery Considerations for Oracle Backup & Recovery
What are Virtual Machines? They re files on disk
Is the Image enough? Traditionally, no
What about Crash Consistent image-level backups? Microsoft VSS helps (application quiescence) but may not be enough
Applications The biggest barrier Traditional image level approach does not address application item-level restore (AD, SQL, Oracle, Exchange, etc.)
What about image plus agent? Solves many issues, but you re backing up your data twice
It s not about backup, it s about recovery Backup is easy, recovery is hard
Considerations for Exchange Image-Level Backup & Recovery
3 rules of Exchange backup Microsoft perspective: In order to be compliant with Exchange Server, VSS based backup applications must follow three basic requirements to ensure the integrity and recoverability of shadow copy backups. If these requirements are not followed, Microsoft... will not be able to troubleshoot backup and restore issues. Rule 1. Exchange must be backed up exclusively through the Exchange VSS Writer. Rule 2. Backup should not be relied on until the backup application has completed integrity verification. Rule 3. Restores to original location must be done exclusively with the Exchange VSS Writer.
Rule 1: VSS-aware backup Veeam implements proprietary VSS integration since 2.0 Fully automated (no agents or manual deployment) Supported directly by Veeam More information: Veeam Backup v5 FAQ: http://tinyurl.com/v5faq
Rule 2: Verify before you rely SureBackup Recovery Verification Great flexibility (supports custom scripts) Choose method of verification that is sufficient for you Remote run eseutil or isinteg on test VM (no stress on production) Log on to test mailbox via HTTPS and query test email message Keep in mind DC dependency! Exchange must see DC to be able to properly boot in the isolated environment.
Rule 3: VSS-aware restore Restores to original location must be done exclusively with the Exchange VSS Writer, and in correct sequence: Boot up Exchange VM with mailbox stores dismounted Tell Exchange VSS Writer to perform restore from VSS snapshot Mount mailbox stores Veeam implements these Microsoft requirements Most image-level backup vendors do not do this, they just boot VM normally like there is no Exchange in it Perform test restore to check your current solution!
Rule 3: VSS-aware restore Look for presence of these events on the restored VMs: Event Type: Information Event Source: MSExchangeIS Event Category: Exchange VSS Writer Event ID: 9620 User: N/A Computer: ServerName.contoso.com General: Exchange VSS Writer (instance GUID) has processed pre-restore events successfully. Event Type: Information Event Source: MSExchangeIS Event Category: Exchange VSS Writer Event ID: 9618 User: N/A Computer: ServerName.contoso.com General: Exchange VSS Writer (instance GUID) has processed post-restore events successfully.
Transaction logs If transaction log files are not pruned after backup, the log files accumulate until they fill all the available disk space. Exchange VSS Writer implements transaction log pruning capabilities. However, VMware Tools VSS is NOT a backup application and cannot know if backup was completed successfully. Thus, it cannot process transaction logs by design. Any application riding on VMware Tools VSS instead of providing specific integration will not truncate logs.
Transaction logs Some solutions do provide transaction log pruning, but perform log pruning right after the snapshot is taken. This approach is actually worse than no pruning: If backup does not complete successfully, you will not have a good backup, and your transaction logs will be gone. You will not be able to restore in case of disaster. Perform test backup to check your current solution! Perform backup, wait for the job to complete successfully, ensure transaction logs are actually pruned. Perform backup, but this time reset the backup server while the job is running (after virtual disk copy starts). Transaction logs should NOT be pruned.
Granular recovery challenges Typically granular recovery from an image-level backup was difficult Restore entire Active Directory and Exchange servers to an isolated environment (resource intensive and time consuming) 3 rd party software that mounts Exchange data store still requires data store to be extracted first (time and disk space), additional cost associated Agent-based solutions could back up the Exchange data, but that s backing up the data twice when working with image-level backups, additional costs for application agents
Granular recovery with vpower Veeam s patent-pending approach fully utilizes the existing virtual infrastructure Veeam application group and virtual lab features automatically create an isolated environment vpower allows you to run the AD and Exchange servers directly from the backup files, no extraction necessary Veeam s Exchange Recovery Wizard utilizes Microsoft Exchange APIs and connects to both the production and isolated environments Exchange item-level recovery in minutes, not hours!
Exchange AIR EXCHANGE AD / DNS EXCHANGE Isolated Sandbox Virtual Lab vsphere vpower Backup Store vpower NFS VMFS datastore
Considerations for Oracle Image-level Backup & Recovery
Oracle on Windows Oracle 11g has a component Oracle VSS Writer Select Windows OCI Components on the Oracle 11g database install wizard Support for Oracle 10g started with patchset 10.2.0.3 and above Requires ARCHIVELOG mode Veeam app-aware processing supports transactionally consistent backups with Oracle VSS Writer installed. No additional settings required (works out of box)
Oracle on non-windows Research among our customers showed 3 equally popular approaches to transaction consistency, from most to least intrusive: Method 1. Performing suspend/resume of database Method 2. Putting all tables in hot backup mode Method 3. Doing crash-consistent backup All methods requires ARCHIVE LOGS mode, NOARCHIVELOGS mode supports cold backups only!
Method 1: Suspend/Resume Effectively closest to cold backups A few seconds downtime (which many smaller shops find acceptable during off hours) Uses VMware pre-freeze and post-thaw scripts: Pre-freeze script executes ALTER SYSTEM SUSPEND; Post-thaw script executes ALTER SYSTEM RESUME; For more info on using pre-freeze and post-thaw scripts with Veeam Backup, see whitepaper below. Hot backup of MySQL on a Linux VM http://www.veeam.com/backup_of_mysql_on_a_linux_vm_2_wpp.pdf
Method 2: Hot Backup Mode This method puts all tables in hot backup mode that is specifically designed for hot backups. No downtime. Oracle ceases to make direct changes to the database file, so it can be safely backed up. Trade-off is increased use of rollback segments, redo logs, archive logs, and internal buffer areas. Virtualization and VMware snapshots are huge win here, as database only stays like that for a few seconds: In legacy file-based backups, you have to size redolog files big enough to survive the entire hot backup window, and account for the extra load from having to take copies of the changed datafile blocks.
Method 2: Hot Backup Mode Uses VMware pre-freeze and post-thaw scripts: Pre-freeze script executes alter tablespace TABLENAME begin backup; Post-thaw script executes alter tablespace TABLENAME end backup; Scripts must enumerate and cycle through all datafiles and place them in hot backup mode. A number of sample scripts are available on Internet depending on your OS.
Method 3: Crash-Consistent Backup Backup Oracle w/out special measures. Be sure VMware Tools quiescence is disabled (beware of SYNC driver!) No downtime, trade-off is having to convince your DBA :) Comments from backup administrators: We used this method for years, and although our DBA was skeptical for a while, I eventually won him over because we wrote a script that actually mounted our storage snapshots on the backup server, started Oracle, preformed the "crash recovery" and ran a dbverify every night. This ran for years without a single recovery failure. during the SAN corruption issue earlier this year that caused us to loose 4+ TB of VM's, we restored almost 1TB of Oracle databases with Veeam with no issues whatsoever. All of these databases were backed up with no pre-freeze/post-thaw scripts, and all restored correctly, performed media recovery, and ran flawlessly.
Method 3: Crash-Consistent Backup Why this works? Virtualization and VMware snapshots make it possible! Oracle works hard to protect data in the event of a system failure: Writes transactions to the log before committing them to the datafile Provides methods for recovery of volumes after a hardware or system failure, or other issue. VMware snapshots are "atomic" and snapshot every VM disk simultaneously. This is no different that if the system had somehow "crashed" at that moment and Oracle is about as safe as you get in that circumstance.
Oracle AIR vpower enables granular Oracle recoveries at no additional charge Restoring tables and individual records: Oracle SQL Developer is great tool and it is free Investigating schema changes and restoring schema Very common issue Oracle admins and developers have to deal with. Tools are readily available such as RedGate Schema Compare for Oracle, but they require connecting running to previous copy of database - vpower make previous copy available for you in just a few minutes then you just compare current and previous copy!
U-AIR No application-specific agents to license No additional products to buy No special backups to create Durable not tied to application internals Supports ANY virtualized application, on ANY OS Also includes wizards for Microsoft Exchange, Active Directory and SQL Server Administrator- or user-directed recovery
U-AIR workflow engine Maintains separation of permissions between backup administrators and application owners Application owners install wizards on their workstations Application owners submit restore requests to backup administrators Backup administrators approve/deny requests Application owners get notified, can then finish the wizard to restore items
Veeam Backup & Replication
Breakthrough technology Publishes the backup as a regular VMDK file Compressed, deduplicated, incremental Runs a VM directly from the backup file Runs on regular backup storage Does not make any changes to the backup file Automatically manages an isolated virtual lab Workflow engine manages requests Proxy appliance provides access into the isolated environment This is why I love Veeam. They take all the virtualization technology pieces on the table and put them together into a functioning business solution. Brett Westover IT Administrator Therapeutic Research Center
Instant VM recovery Fast: Start VM directly from backup file Readily available: Uses existing backups and backup storage Buys you time: Users keep working while you troubleshoot the problem
Instant VM recovery Storage vmotion / Veeam Backup & Replication
How it works Publishes the contents of the backup file as a datastore that an ESX(i) host can connect to Uses vpower NFS server Works with compressed, deduplicated and incremental backups Use Storage vmotion, replication or hot VM copy to complete the recovery
But wait, there s more! Happy Holidays from Veeam! Announcing our NFR for VCP, VCI and vexpert program Our thanks to the VMware community 2 socket NFR license for: Veeam Backup & Replication with vpower Veeam ONE Solution (Veeam Monitor, Reporter and Business View) http://www.veeam.com/go/free-nfr-license
Thank you! For more information Contact the presenter: Email: doug.hazelman@veeam.com Twitter: @VMDoug Go to www.veeam.com/vpower Veeam Blog www.veeam.com/blog Veeam Forums FAQ: http://tinyurl.com/v5faq Follow @Veeam on Twitter Contact your local Veeam ProPartner or a Veeam sales representative: www.veeam.com/contacts