snapview question

Question

how many reserve LUN's(what sizes) i need to accomodate a snapview for about 5 TB of data .
Here is what i have done .
I have assigned 2 500 GB LUNS to reserved SPA and SPB .
I have created a snapshot of 600 GB LUN . No problem. I was able to start session , activate , de-activate , etc
I tried to create another snapshot of 500 GB LUN , No problem . However when i tried to create a session for that , it complaint about not having enought space .

To my understanding , we just need around 10% of actual size of the LUN size . This means for 5 TB , with 1 session running for each snapshot , 1 TB should be plenty .

Also , what should be use - persistent or consistent .

Any help appreciated .

dseth1 · Answer

i got the white paper and now going throught it . I have removed the big size 500 gb luns used for reserved pool . i have created 50 GB - 20 smaller luns are will try it shortly . we are on flare code 22 .
i have about 10-15 luns of different sizes - ranging from 50 - 600 GB . Let me try it smaller reserved pool and see how it behaves.
thanks for quick reply .

calle2 · Answer

That's a lot of questions...but the answer is it depends

For example:
- You can't "guarantee" which source LUN gets what RLP
- 10% is quit low unless your data doesn't change very much or has high locality.
- EMC recommends 20% and if you use the wiz then it will try to grab 30%...
- If you got very random data then possibly every write to the source could cause a 64KB write to the RLP (i.e. IOPS x 64KB x time).
- Writes to the snapshot will add a lot more to the above calculation!
- As of FLARE 24 all sessions are persistent.

To get a better understanding of how RLPs work I would highly recommend you to search for the following document on the EMC knowledge base "White Paper: EMC CLARiiON Reserved LUN Pool Configuration Considerations - Best Practices Planning".

I hope this helps
Carl

RRR · Answer

I'm curious here too, since I've seen errors like the ones ytou described. RLP was not in use at all (config'd with 24 LUNs though) and when I was trying to create a snapshot on a 350GB LUN (can't remember exactly), Navisphere complained about not being able to do it. I didn't even started and activated: it was all the way in the beginning while creating the snapshot !

calle2 · Answer

I am sure you know this already but creating a session allocates a RLP and it could be MV/A or ISC instead of SnapView. Also the RLP must be at least 0.02% of the source LUN size.

RRR · Answer

Also the RLP must be at least 0.02% of the source LUN size.

So 0.02% of 350GB is like 0.7GB. I'm pretty sure my RLP was something like 20 x 1 or even 2 GB, so 20 or even 40GB in total.
Do you really mean 0.02% ? Or 2% ? And is this number per RLP LUN or for the whole RLP cumulative ? And in Flare 19, does this count per RLP per SP or the whole RLP of both SP's together ?

calle2 · Answer

Please don¿t shoot the messenger

The 0.02%, and yes I mean 0.02%, is mentioned in the ¿EMC CLARiiON Reserved LUN Pool Configuration Considerations - Best Practices Planning¿ white paper.

Some more basic info:

In FLARE 24 the SnapView reserved LUN pool is per storage system as opposed to per storage processor. A reserved LUN added to either storage processor will be available for allocation to either SP, as long it meets all criteria¿s including size... In FLARE 19 the same but obviously no trespass¿

How it would work in even older FLARE releases that used a pool area of snapshot cache that was allocated to each storage processor such that all running SnapView sessions on one storage processor would use the same allocated cache LUNs¿doesn¿t bare to think about!

But I guess what you are asking if a second RLP is allocated to the same source is the minimum size requirement again 0.02%... Honestly I don¿t know but taking a wild guess I would imagine so. For sure it will be for the first and therefore the cumulative RLP count per host.

But to really twist the scenario what if the source LUN was expanded¿

calle2 · Answer

Cool! But plz plz come back and let the forum know of your findings...

RRR · Answer

Trying a DDOS with giant pigeons dropping cars on our data center ?does anyone else have any great ideas ?

RRR · Answer

Sure, but not everything we'll test will probably be suitable for on the forum.
We're thinking about:
- pulling disks
- adding hosts to storage groups that have no business there
- pulling hosts from SGs while in production
- removing zones on a switch
- enforcing standby zonesets instead of the active ones
- pulling ISL's
- switching off SP's
- switching off power of switches and arrays

...any other ideas ?

RRR · Answer

But to really twist the scenario what if the source LUN was expanded I'm going to set up a list of things to test ! This will be one of them.

dynamox · Answer

setting off water sprinklers ?

CLARiiON

snapview question

Was this post helpful?