Need some advise on migration strategy to migrate Solaris based physical clients from one DC to another. The target DC is a different NetBackup domain.
I am planning to use NetBackup BMR (currently NBU is being used in source DC for client backup with NO BMR). I understand that a BMR master server and Boot Server (seperate boot server) needs to be introduced. The SRT process
The target DC will have NBU Master/Media servers and a compatible storage library.
Solved! Go to Solution.
BMR recovery for host running Oracle RAC is not been officially qualified.
Generally to protect whole system running Oracle database, two policies are recommened. One is BMR to recover OS + APP installation back and other one it NB ORacle APP agent based backup to ensure database restored is consistent. App backup will be more frequent than BMR backup to make sure db point-in-time backup is wide.
So using this two backups, first recover system using BMR (even OS+other file systems as well). Then recover Oracle database from the other agent based backup. You will get the whole server as it is on target.
Hi Itegral and Marianne,
Actually BMR can be best used to protect your clients on primary domain to disaster recovery domain by leveraging Auto Image Replication feature introduced in 7.5 release.
In case you have pre-7.5 environment then,
a. you need to first import BMR enabled backup image at your DR NB master.
b. setup bmr master server
c. from the imported backup image, recover only bundle.dat file availalbe in /usr/openv/netbackup/client/data folder in that backup. Use NB recovery agent to do this.
d. Using bmrs -import ..command import this BMR information in DR BMR master server.
e. You need to make copy of this imported BMR client config and edit it to change nb master, media host enties matching your DR environment.
Basically few manual steps are required to be done in case you are importing backup manually.
Now in case you have AIR+BMR enabled at primary domain then you dont need to do above steps manually. AIR replication and import will happen automatically for your primary domain protected clients backup. Also BMR configuration will automatically imported on your DR BMR master server. Obviously you need to make sure you have enabled BMR master server onto DR NB master server.
Now if recovey is intended then you need to setup BMR boot server as well in DR domain; create matching SRT onto it. You are done. You can recovery your primary domain client onto DR domain.
Both of these mechanisms have been tried out successfully by BMR customers.
Mandar - thank you for the detailed response.
Will it help if I move the NetBackup (Master/Media on same server) to the target DC?
The new DC is isolated from the old DC. I can perhaps consider using the same OS, version, patch level, architecture, logical volumes, installation path and storage LUNs then recover the entire catalog files using the DR wizard - http://www.symantec.com/docs/TECH77448
I believe in that case I dont have to go through the manual process.
Now the question are; since the BMR backup for client needs to be done in the source DC, do I need boot server in the source DC? where does the SRT process (prepare to restore) need to run, source or target? Do I still need to import the backup images from tape?
1. With the process you mentioned, you will still need to do manual steps for customizing BMR config to match your new DC master, media name-ip-addressess. Refer BMR admin guide, "Managing client and configurations" to know how to change host details in BMR configuration.
2. You are right. You do not need to have boot server in the source DC as you are just backing the clients there. But in case you need DR support at source DC as well then you need to setup boot server/s as per need.
3. PTR needs to be run at target as there you are recovering the client.
4. regarding tape import: I am not sure if nbcatsync will manage the catalog tuning automatically. Else you need to import it from tape. You need to check that out. I believe you may not require import if catalog is there.
Mander, please excuse my lack of understanding on solaris and BMR feature.
I will NOT be making any changes in the hostname/IP address for Netbackup Client and Master/Media server - then why would the manual steps be required?
Also, we also intend to move data center to different physical location (source DC to target DC).
1 - The new servers in target DC (near similar to the source DC) will not have solaris OS pre-installed;
a) what is the best way to approach this kind of BMR restore when new servers in target DC have no OS?
b) The source servers have SAN volume (LUNs) attached - will BMR process take backup of the SAN volumes? and how will it restore the backup of those volumes on the above servers (no OS pre-installed) and no SAN presentation (zoning etc)?
c) Would it help restore the BMR and associated SAN volumes if I install the OS on new server hardware before I initiate a BMR restore process?
d) I have business critical applications (mainly Oracle) running on those servers and data/database on SAN volumes (LUNs). The SAN in the target DC have the same configuration as source. Are these applications expected to run without issues after the restore without breaking binary relationship with data/database? How can I make sure that application recovery is smooth.
Profuse of thanks for your advise.
>> I will NOT be making any changes in the hostname/IP address for Netbackup Client and Master/Media server - then why would the manual steps be required?
I consider your DR/target site has different NB master, media server where you are importing backup images. Isnt it? In that case your server names/ip details will be different? If you are just moving NB servers in primary domain to target then you will not require any bmr config manipulation.
>> a) what is the best way to approach this kind of BMR restore when new servers in target DC have no OS?
I consider that you mean servers where you want to recover your primary clients. Correct?
You need to discover them using BMR SRT first. This process captures hardware details like NW, disks etc on target bare metal HW. Then you need to map primary clietn config to this discovered one and follow futher recovery operation. BMR admin guide best explains this.
b. Yes, BMR backs up SAN volume data as well if file systems on top of them are mounted over client and NB backup policy has ALL_LOCAL_DRIVES set in backup selection list.
After BMR target hardware discovery process you can map all original disk/luns to newly available storage and do recovery. The storage in target can be different than original; hence BMR disk mapping helps to customize original file systems/ volume to match target HW storage.
c. You do not require to install anything on dest HW if you are doing BMR. You get primary client recovered on target as it was BMR backed-up. :-)
d. I do not anticipate any problem to recovered applications. Apps should not bother about underneath storage HW change. BMR will be recovering file systems as original to there should not be any impact to apps whether you recover them on SAN LUNS or local hard disks on target.
The target DC is not a "DR". Its a new data centre.
Its a DC migration activity I am refering to. We are migrating servers, not moving servers physically. I'll be installing a new NBU Master/Media server in target DC keeping the same hostname/IP address and restoring the catalog. This should allow me run the restores on target DC with no issues as we are keeping the same hostname/IP address for all machines including clients.
We are migrating from DC "A" to DC "B". The source DC (DC A) is hosting our clients servers. We are migrating the data to target DC (DC B) but not taking the physical machines (hardware) with us from source DC (DC A). We will have near similar hardware in target DC (DC B). Becasue both the DCs are not connected with each other, we will have to move the hardware config using BMR. In target DC (DC B), we will be keeing the same hostnames/IP address and SAN volumes etc.
New servers = the hardware (new client) I want to restore my "original client" to (usin BMR restore). This hardware will be "near similar" to the hardware we have in our source DC. This hardware (servers) will not have any OS and obviously no SAN volume config on it.
Clarifying my questions inline with above:
a) I want to take BMR backup of clients in DC A and restore it on another server (similar hardware) in DC B. The server in DC B does not have any OS on it.
Q: With no OS on the server in "DC B", how will I discover it using BMR SRT?
b) BMR backs up SAN volume data as well if file systems on top of them are mounted over client and NB backup policy has ALL_LOCAL_DRIVES set in backup selection list.
Q: Does it mean, I'll have to install OS on new hardware in "DC B", mount the volumes on this servers and then run BMR restore process?
c) With no pre-installed OS on new client and no SAN volume zones to new client, how the discovery process will show me the target mappings (new SAN volumes)?
Q: Do I have to turn the new client ON (with no OS installed) for BMR target hardware discovery process?
Shall I zone the SAN volumes to this new client so the discovery process could map all original disk/luns to newly available storage?
I am quite confused mate, please help!!
Mander, could you assist please?
Q1: With no OS on the target client server in "DC B", how will I discover it using BMR SRT? do I have to turn the client on?
Q2: I understand that post discovery process allows volume/lun mapping. With no OS on the target server, do I have to zone the storage LUNs to the new client for it discover the new LUNs?
After that I believe I have to create new client configuration and run PTR?
thanks for your help in advance.
Q1. You need to create BMR SRT matching your client OS. You need to setup BMR Boot server and create this SRT on it. You can either boot your bare metal client over NW or over media SRT for discovery.
You can refer below links which explain Windows 2012 R2 physical machine recovery onto a bare virtual machine. Though this is done in the same domain; it would give you understanding of BMR recovery process onto different system.
In case of UNIX system, make sure destination hardware if of same type like original.
Q2. You may want to attach SAN Luns physically to your target. You do not require to configure them and BMR recovery should take care of it.
In case intended, you can do recovery of OS volumes only using BMR and later manually configure data volumes over SAN.
the guide says; on Solaris 10 BMR automatically restores the SAN-attached data volumes if HBA drivers are available in the restore environment.
Q1 - So if I understood it correctly, present (zone) the SAN luns to the target client HBA and run the restore. It should automatically pick the drive letters (E:\ or F:\) and restore the data to it?
Q2 - If I do recovery of OS volumes only on Solaris then manually mask the luns, then restore the data manually on luns. Will the application hosted on OS volume pick up the data automatically after reboot?
ANS to Q1: Yes. After BMR discovery of the machine having SAN Luns connected to it; if captured BMR configurations shows those SAN Luns then that means it can restore file systems on top of it during recovery.
ANS to Q2: That should work. Only additional manual change require would be verification of /etc/vfstab file where mount points are defined. In case user is recovering only OS using BMR then BMR automatically marks non-OS mountpoints as commented. This is for safety purpose and user can uncomment those entries later manually.
This is perfectly fine and works.
Unfortunately I do not have any whitepaper kind of stuff on this specific points. You can find BMR admin guide useful to know stuff around this process.
No technote available as such. But you can find related information in BMR admin guide. Especially setup/unsetup bmr in cluster and info about recovery in cluster.
To give you brief about BMR protection to cluster nodes:
BMR does recovery of individual nodes and does not configure any cluster settings over node post-recovery. It is expected that the node automatically joins cluster once it is recovered back. Though user may need to do cluster specific configuration further.
Also BMR today understands VCS and MSCS and marks shared disks restricted in that nodes bmr config. This prevents any shared disks re-write/restore during that nodes bmr recovery. Though it is recommended that before doing bmr node recovery, confirm disks restriction using bmr config admin gui wizard. user can do intended changes here. You can find note in admin guide related to this.
Thanks for keep responding Mander, I really appreciate!!
Regarding restoring the primary Solaris client to an identical hardware in new data centre, do I still have to follow steps for Unix http://www.symantec.com/docs/TECH62678???
Where does the "creating Boot Media" process happen, on BMR Master server?
Yes, you need to follow this technote. Especially in case where your target system is different (e.g. new/old) than original. Anyway as i mentioned earlier target hw (which is bare metal hw) disocvery using BMR SRT is required to understand its NW MAC, hard disks info. You need to initialize your original client bmr config with this discovered one -> do required mapping and recover.
"Creating boot media" process happens over BMR Boot server component. Dont worry BMR Boot server is not a big server but a lighweitht component which prepares BMR Shared Resource Tree (SRT). SRT contains mini-OS + NBU client; which stages a temporary recovery environment to recover your client. Your bare metal hardware boots over this SRT for recovery or even to do machine discovery.
BMR supports NW as well as media based recovery. User needs to create SRT to match their client OS and version. bmrsrtadm wizard on bootserver does this. In case user wants to recover clietn over media then user needs to create SRT media which is "creating boot media". bmrsrtadm create SRT ISO which can be burned over dvd/cd.
I would recommend you to go through bmr admin guide to know all this in details.
If multiple boot servers are required, can these be run in Solaris Containers/Zones to allow use of one additional server?
Is there any special instruction about SUN SPARC Ent servers?
Any recommendation on multiple network interfaces (VLAN) i.e. Application to DB, backup, ILO, Load balancer to App server etc.?
>> If multiple boot servers are required, can these be run in Solaris Containers/Zones to allow use of one additional server?
BMR works fine in VLAN setup where a single bmr boot server can be used to provide NW boot based recovery for clients located in different vlans. Here you need to configure your switch settings to forward DHCP boot requests to the boot server.
Q1 - Is Solaris cluster running Oracle RAC supported by BMR? We have NetBackup 7.01.
Q2 - If not then what alternative can you advise in the server migration situation where I intend to restore BMR of the source client to the target client (server)?