Local Directory Container Storage pool protection options

ILCattivo

ADSM.ORG Senior Member
Joined
Jul 9, 2013
Messages
192
Reaction score
14
Points
0
Location
Oxford, United Kingdom
PREDATAR Control23

I have a business requirement to protect the same storage pool data on two different types of media.

With Spectrum Protect 8.1.x Server the most common way to protect a container storage pool is by utilizing the 'Protect stg' cmd followed by the 'Replicate node' cmd for raw and meta data protection to a DR site Spec Pro Server over a decent WAN link.

I have a customer who would like a 2nd level to this disk to disk protection by backing that same container storage pool & the system db backups to a Fibre Channel attached Tape library at the Primary site. (Ransomware gitters I believe)

Anyone have this scenario setup in place on this level of Spectrum Protect and how is it best configured in terms of maintenance scripts, as I understand it you cannot run two 'Protect stg pool' cmds concurrently.

Thanks
 
PREDATAR Control23

I have a customer who would like a 2nd level to this disk to disk protection by backing that same container storage pool & the system db backups to a Fibre Channel attached Tape library at the Primary site. (Ransomware gitters I believe)

So I've yet had the ability to stand up replication server... However jumped into container pools for the inline deduplication and better compression they provide. I do have my container pool set to use a local copy. On the local copy I have 5 (IBM I think recommends 6 - I just don't have the tape resources currently) protection processed defined.

The really nice thing about sending the container data to tape is it compressed/deduplicated so it copies pretty fast. Obviously that initial protect will take a while depending on size, and you'll need to plan accordingly.

Personally I'd just look at running protect stgpool type=replserverwith a wait statement. Next line in your script would be same process with type local. Again, you could add a wait to it if need be.

Something else to add: I'm on 7.1.7 and as such have no control over the tape reclamation process of the protect storage pool. For example 558GB of data protected took 1 hour 58 minutes. This is on LTO6 drives in a 3584. Today, 611GB of data, along with 490GB of data movement (reclamation process) took 4 hours and 22 minutes.
With SP 8.1.x you have the ability to schedule container copy reclamation outside of the protect process. If tape drive resources are limited it may be something else to think about. That's one of the reasons I'm pushing hard to get to 8.1.x but some other items holding me back currently.

Oh and make sure your customer understands that recovery from the container-copy pool could be time consuming. http://www-01.ibm.com/support/docview.wss?uid=swg27048653 Guidelines for using container-copy

Hope it at least gives you some ideas. And if you don't mind let me know what you end up doing? I have a feeling we'll be doing the same thing once we are to the point where we have a replication server spun up.
 
PREDATAR Control23

Thanks for the reply RO. That all makes sense.

The good news here is that we will be in a position to do a dry run or POC of this scenario. The bad news is that it has to be somewhat of a 'big bang' approach in terms of initial ingestion into the new Spectrum Protect 8.1 Server, and that's the part that concerns me. Normally I like to phase clients gradually from one backup server to another, but as this is (in my particular circumstance) a Data Centre move from one geographical location to another it can't be done that way.

The current server is at level 7.1.500 and sits at SITE A with a TS3200 attached which receives FILE DevClass Storage Pool backups via the BACKUP STGPOOL cmd.

SITE A will be replaced with SITE C whereby there'll be a Spectrum Protect 8.1.x Server waiting for the new ingestion of data to come over from SITE A in the form of IBMv7000 which will hold all the VMWare data stores and vm's previously located at SITE A.

SITE A will then protect it's dir container Storage Pools to..
1. SITE B to disk > Spectrum Protect 8.1.x dir container stg pool.
2. locally to tape via FC on LTO Ultrium 7 to give it the 750MB/s r/w speed

Primary recovery of STG Pool data should always come from SITE B unless compromised whereby we then have the secondary but slower recovery option of the same protected data & db backups locally on TAPE.

This is still all in the planning stage so will keep you updated on how things go..

Cheers
 
PREDATAR Control23

locally to tape via FC on LTO Ultrium 7 to give it the 750MB/s r/w speed
Off topic here I know but NICE!
I'm having a heck of a hard time of pushing my LTO6 drives to even anything close to their max rate. As you can see 558GB I posted above is only averaging out to 113MB/s and that's spread over 5 drives! So little over 22MB/sec per drive. Likely a little bit more due to mount times etc but still... Don't get me started about deduplicated file pools to tape...All disk storage pools backed by a V5K. One of these days I'll get about to posting here for performance tips...

Ok back on topic:
The plan sounds really elegant. Sadly I have no experience with the replication/protection to another server.

That is pretty much it! Good luck!
 
PREDATAR Control23

Blimey..!! if I get those kinds of r/w speeds for Tape when we move all this data to the new DC for stg pool protection then I'm gonna be in big trouble!!

I'll be honest, I'm no tape expert myself, but if you don't get any joy from this forum then if I were you i'd get an IBM PMR raised for that, cause there's deffo something not right there! LTO U 6 should be giving nearer 400MB/s per drive.

Good luck.
 
PREDATAR Control23

Hi,

To evaluate performance for lto drives, you have to look at the entire chain. DISK(SSD/15K/NL) - connectivity(SAS/FC/iSCSI) - switches(san/network) - Controller (sas/fc/nic) - drivers - os - application - os - hba - lto6

Anywhere in this chain, you can find parameters to tune, drivers to update and so on.

In addition, the type of data pushed (small files/big files) will have a huge impact on perfomance. One 8 Gb HBA can push four lto6 drives if the data is uncompressed (lto6 has 160 MByte/sec max speed). If the data is compressed, you can have maybe eight drives on one HBA. And, do not use tapedrive and diskdrives on the same HBA.

There are some trace flags that can be used to find where the bottle neck is as seen from Spectrum Protect.
 
PREDATAR Control23

(Not to hijack the thread but to divert it on another road for a moment)
Trident,
There's some minor cable changes that I can do at the physical layer. In about three weeks time I'll be making those changes which should help I hope. But beyond that its work with what we have. Figure once I get all that done I'll make a new thread about my current setup and recommendations the community here has.
 
Top