Adding new drives enclosures to the EVA8000

[Nahimm]

There is a EVA8000 system: 6 shelves/70 disks.
According to the plan it is required to enlarge up to 10 shelves/120 disks.
We are interested in the following issues:
- Is it possible to find out the quantity of disks in RSS in the operating system (with 70 disks)?
- Is it possible to reduce/enlarge the quantity of disks in RSS in the operating system (with 70 disks)?
- How to add 50 disks to the operating system correctly?

[Steven Clementi · Accepted solution]

- Is it possible to find out the quantity of disks in RSS in the operating system (with 70 disks)?

yes. But let it be known that it will look like this...

8 RSS's of 8 disks each and 1 RSS of 6 disks.


- Is it possible to reduce/enlarge the quantity of disks in RSS in the operating system (with 70 disks)?

The EVA will automatically adjust your RSS's according to the amount of disks you add. You have no real control over this.


- How to add 50 disks to the operating system correctly?

Before adding in the disk shelves... power them up and populate them with all of the drives. After everything has been powered up for a while, connect the first shelf to the backend network. it will probably do a code load on the EMU. Wait for it to finish.

Add the next shelf, rinse and repeat.

Always give ample time for the shelves to get updated and situation properly. Always make sure the shelf is visible in Command View without any errors before moving on to the next shelf.


With 120 disks, you will end up with 15 RSS's of 8 disks each.


Steven

[Timothy Cusson]

I agree with Steven except for adding the disk enclosure with all the dirves installed. If the EVA is running adding 14 drives at once will cause some of the drives to receive improper or duplicate disk numbers.

I would power up the disk enclosure with only one disk in slot 1.
Connect the enclosure ID cable.
Connect the A side fibre cable.
Connect the B side fibre cable.
Verify the EMU has proper shelf ID number.
Verify the one disk appears in the ungrouped disk group.

Repeat for the other enclosures.

Next, add one disk at a time and WAIT at least 1 minute, check drive shows in Ungrouped Group.
Repaest until all disks are added.

Alternate methode.

Shutdown the EVA, connect all the shelves, install all the drives, power up EVA, verify each shelf has proper ID number, verify every I/O module has proper firmware, verify all new disk are in ungrouped group.

For both methods run disk drive firmwae bundle.

[Steven Clementi]

"If the EVA is running adding 14 drives at once will cause some of the drives to receive improper or duplicate disk numbers.
"

Yes, that is a possiblity. I have not seen it happen as of yet unless there were already existing comm problems.

Always check your EVA out before doing ANY work to it. Make sure CV is reporting all is good. Check the hardware for amber lights, etc..


Steven

[Víctor Cespón]

Let me clarify two points:

The operating system on the server does not see the disks, much less the RSS. And does not need to know anything about the internal structure of the EVA.

Each disk is connected to two loops, and the adresses on each loop are fixed by the position (enclosure/bay) of each disk. The will be no conflicts unless you connect something on the wrong place.

[Timothy Cusson]

Reply To Steven re disk numbers:

I've seen EVA drives with no disk name-number, i.e. disk42, disk92. This is usually caused by adding more than one disk at a time and not waiting one minute between additions.
To correct this problem I upgroup the disk, remove the disk, re-insert the disk.

[Steven Clementi]

"...usually caused by adding more than one disk at a time and not waiting one minute between additions."

Right, exactly.

I've never seen it happen when you add a whole shelf of disks though mainly because all the disks are already spinning up and initilaized.

When adding in single disks, you certainly need to do one at a time to let it spin up, initialize and join the array.


Steven

[Timothy Cusson]

Reply to vcespon:

Let me clarify two points:

The operating system on the server does not see the disks, much less the RSS. And does not need to know anything about the internal structure of the EVA.
>>> I agree.

Each disk is connected to two loops,
>>> I agree.

and the adresses on each loop are fixed by the position (enclosure/bay) of each disk.
>>> I disagree. The disk name-number and the disk ALPA number are NOT related to physical location. Upon disk insertion, the disk does its power on self test, and then it requests a disk name-number and an Arbitrated Loop Physical Address. The EVA "gives" out these settings based on a confidential internal algorithm somewhat like DHCP.

There will be no conflicts unless you connect something on the wrong place.
>>> I disagree. The HP best practices clearly state you must wait at least one minute between online insertions of disk drives. The reason is to avoid possible disk name-number conflicts or ALPA number conflicts.

[Timothy Cusson]

Reply to Steven:

I've never seen it happen when you add a whole shelf of disks though mainly because all the disks are already spinning up and initilaized.

>>> I've never tried that method. It would save alot of time.

Tim

[Víctor Cespón]

Tim, on the EVAs 4000/6000/8000 the AL_PA addresses are fixed, there's a table that tells you what AL_PA will it get depending on the enclosure and bay.

IF you connect something on the wrong place, like a disk enclosure on ID 7, or a disk on enclosure 17 bay 13, it will get a soft AL_PA, since there's no legal AL_PA for that combination.

The name of the disk you see in Command View is completly irrelevant for the EVA (like all other names). The EVA uses only WWIDs, there's a table (MLD) relating WWIDs to friendly names (you must get a copy of it when using EVAperf for example).

Having two disks with the same friendly name can confuse you, but not the EVA. It happens sometimes and you can just rename it to whatever you want, it does not need to be Disk###.