Bets method to safely decommission VMAX Storage

Hi AMC,

i am not quite sure i am understanding your question, but there are some methods to "recognise" which device you want to move or not.

First of all you can use the symstat command for the specific device to check if you have I/O or not.

It is not of course safest method but one more tip if you don't know it.

But don't count on this 100%.

Of course it depends on the O.S.

If you are in HPUX, you can use the INQ tool, executing it tells you all the devices attached to your host and the EMC ID.

By this tool you can have the EMC IDs for all attached devices on the host.

There is also INQ for all the O.S.

Also everytime you create a device you can identify "set identifier" on each device, so that will help you in order to distinguish the devices.

Also if in HPUX you are using Veritas, with vxdisk list and INQ you are 100% SAFE and SECURE which device you are going to remove from the host.

Using symstat and\or INQ could be the answer or at least part of it so thank you

We have a multi OS\Array environment and we don't always get great info back from the Platfrom guys so what I'm basically looking for is a way to second-guess things.

We have had a scenario where we were asked to remov a LUN and it was identified by it's decimal LUN ID which is fine except the Server had LUN's from 2 different Arrays... you can guess what happened!

yes, we need your expertise

That's exactly what I was hinting at

It's easier now to use the identifiers at implementation time too as you can now add the label when creating devices.  It's in the cli and also in unisphere too, although you have to look for it under advanced (no idea why it's advanced).

From the CLI 7.6

create dev count =10 config=tdev size=240GB ,device_name=OracleData, number= ;

The number will get incremented from whatever you specify, so if I specify 11 the devices created will be OracleData11 to 20;  alternatively you can add the symdev on there instead.

I agree with Zikas, you need to rely on host based tools to determine which logical device on the host needs to be removed first before its corresponding symdev gets removed on the array side. Some platforms have native tools that can display device WWN or device address, on some (like windows) need a little help with tools like inq or syminq.

Hi AMC,

i believe that what you are doing it is like a Russian Roulette!!!

Please don't that again!!!!

Please tell me what kind od O.S. you are using and we will try to help in order to use a safe way to remove the devs.

Are you using PowerPath?

Through PP you can know the EMC IDs and the WWNs of every device.

Also if you are using ESX, through ESX you have the NAA which will help you to identify the device you want to remove!!!!

If you are using ESX you can install VSI Viewer, which tool will help you incredibly!!!!!!!!!

We are here to HELP YOU!!!!! any time!!!!

Use the symaccess commands in order to check which devices are masked, which devices are in masking view, are in Initiator Groups and it is also a double-check for you!!

"Use us" to help you!!!!

Agree with both Dynamox and NY Yankees, host based tools are your best option. Confirm, Check and confirm again with host tools.  Last thing you want to do is accidentally remove a volume in use.

In addition  you can adopt the practice to label your devices in the VMAX.  symconfigure -sid xxx set dev XXX device_name='easy read label';  

This label is then visible though inq utility with the inq -identifier device_name switch.

The same label is clearly displayed beside the device in all views in Unisphere for VMAX and in symdev list -identifier device_name commands. It's an easy way to ensure the systems admin and storage admin are talking the same speak!

NY Yankees wrote: Dynamox, it is going to take a long time to reach you at your level. thanks for your kind words.

no one knows everything, we all learn from each other

Hmmm, yes interesting. I have used the device_name identifier before but hadn't thought along those lines.

I'm thinking something like this..

1) Syminq on host - free and available for most OS's.

2) Add label to LUN:

symconfigure -sid xxx set dev XXX device_name='I'm going to delete you!';

3) OS guys runs inq -identifier device_name

4) Confirmation that is the correct LUN.

5) Optional symstat just to be safe

6) Not_ready device (anyone think of anything better than this?)

7) Leave for a specific period of time.

8) Decomission.

It would make sense to use the device_name identifier at implemetation time so we already know 'whats what' but I am talking about a process for an existing environment with hundreds of servers and thousands of symmdevs so wouldn't be easy to retro-fit!

Dynamox,

it is going to take a long time to reach you at your level.

thanks for your kind words.

Hello Dynamox,

it's been a long time but i will try to be more consistent since today in ECN.

AMC,

Paul is definitely RIGHT!!!

Set the device identification at the begining of every device creation.

Dynamox, absolutely they can be renamed at any time. But for the time they are in use it’s great to know exactly what they are in use for. We’re usually implementing for long term and not tearing up and down the same devices all the time. But I appreciate in a more dynamic shop this could create a headache to maintain.

amccluskey wrote: Thank you guys this is such helpful stuff. This is actually my first post but will definietly be back! Thanks again p.s love this: create dev count =10 config=tdev size=240GB ,device_name=OracleData, number= ;

i am not sure how provisioning works in your shop but when we reclaim device we don't typically delete them, we re-use them for other hosts. By naming devices you will be force to delete them after you decomm them from a host, unless there is a command that will allow you to rename devices. To me that just adds another layer of unnecessary management.