EVA4400: how fault tolerant it is ?; Difference with EVA6X00/8X00

[damiri]

I do small business comparing to most of you and we are often have only one disk group on EV4400. I am interested in your option how fault tolerant EVA4400 is with one disk group ... I know 2 x controller, 2 x power supplies but is there something what can break and bring whole thing down ?

Besides know difference stated on the web, what is really difference between series of EVA in terms of fault tolerance ?

damiri

[Víctor Cespón]

The basic philosophy is the same: dual-port drives connected to two loops, each loop having both controllers in it.
So there's no single point of failure. The M6412 disk enclosure uses CTS switches, so it's faster and better at insolating problems than an EVA4000 and more similar to an EVA6100.

Of course, if going for availability, it's better to have more than one disk group. But the probability of 2 disks failing on the same RSS in a few hours is very low.
And with the new 09500000 firmware you can create RAID 6 vdisks.

[damiri]

So there is not single point of failure with EVA4400?

[IBaltay]

Hi,
EVA 4400 is designed for no single point of failure:
http://h18000.www1.hp.com/products/quickspecs/12893_div/12893_div.pdf

[damiri]

Ok, this is marketing talk. I would like to get real field experience. We are selling this EVA like fault-tolerant but recently we had some issues where things were not fault tolerant...

[Rob Leadbeater]

Hi,

If you are reselling the EVA then I would expect you to know how best to set it up.

Yes the EVA is fault tolerant, but like any system, if set up incorrectly, or used incorrectly then things can go wrong.

Can you share the specifics of the problems your customers have had ?

I'll take a guess that they instigated from not quickly replacing failed disks.

Cheers,

Rob

[damiri]

So far the only issue is with embedded CV. Too slow or not responding at all but this is not big issue so far. We don't have issue with EVA4400 yet, but I have heard few stories and I would like to gather as much info as I can, so we can prepare or recommend something to our customers. My question was mostly related to single point of failure: to get some info on internal design and is this internal design fault tolerant... That's all ...

[Víctor Cespón · Accepted solution]

It's not less than any other EVA. I've seen a single disk bring down an EVA8000 because it starts causing packet loses on both loops and the controllers lose communication with several disks. That's inherent to the FC-AL protocol, the traffic for a device passes through all other devices, so a disk can interrupt the traffic to others.

Theoretically the I/O module or the controller should detect this and bypass that disk, but depending on the controller and disk firmware this does not happen always and sometimes the thing ends in a lot of LIPs sent to all loops and all communication lost.

This is less probable on an EVAx100 or EVAx400, since the I/O modules are now switches, and insolate the traffic for each disk (or that's the idea).

In the EVA4400 the most important thing is to keep all firmwares updated. There are a lot of issues updated and corrected since the initial launch.

Firmware 09500000 is already on the FTP and has many fixes and new features.

[damiri]

Ok, as always, you are helpful. I have heard stories with having one disk failed to bring whole thing down. Especially if you have only one disk group with 8 drives.

Can you send me link for new 09500000 and is there updated firmware for WOCP with CV since this is almost not usable at all ?

[Víctor Cespón]

Release notes and firmware update guide for 09500000 have been published, but there's no official announcement of the firmware yet.

The management module firmware will be updated too to contain Command View 9 instead of 8.1.

[damiri]

I hope they will fix WOCP and CV within it ...

[Uwe Zessin]

CV-EVA V9.0 and RSM V5.0 are available:

http://h20293.www2.hp.com/portal/swdepot/displayProductInfo.do?productNumber=CommandViewEVA9.0


(In my opinion, the WOCP is only suitable for some emergency management tasks and is *NOT* a replacement for a management server, because important things like call home or online disk drive firmware updates are not available)

[damiri]

Thanks, I wasn't aware of it. Where I can find info on HDD firmwares available as well as procedure how to upgrade drives ? I looked under support for drives, but nothing is available ....

[CLEB]

Customer Advisory for XCS 09006000, which also has links to firmware for HDD and WOCP.

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&taskId=110&prodSeriesId=471497&prodTypeId=18964&objectID=c01625532

[James Ackerly]

Hi All,

I have just been involved with an EVA4400 install. We noted that with the EVA 4400 you only seem to be able to shut down both controllers at the same time and not individually.

Also, the controllers aren't labelled as hotswap.

I take this to mean that if we have a controller failure we will need to schedule downtime for the whole eva to install the replacement controller?

We did some basic testing as we are not at the production stage yet. Basically we have a couple of disk groups and published vdisks. We pulled power supplies/ batteries / fans / loop cables / etc, Command View seems slightly more responsive than older versions but still quite slow and often not displaying the fault for quite some time.

We also disconnected all loop ports and device ports from Controller B to try and simulate a Controller failure. Command View basically froze up and was unresponsive. We have tested the MPIO drivers since on our management host and they are working correctly. Has anyone else tested and experienced anything like this?

Cheers,

James

[Víctor Cespón]

Are you using the latest firmware (09500000) and Command View (9.0)?

Comand View only talks to the master controller, if you remove the FC cables from that one, Command View will not be able to access the EVA.
It's not the same a controller with the ports disconneted that a failed controller.

Also, EVA4400 can be shut down individually, in fact it's stated on the controller replacement instructions:

a. In the navigation pane, select Storage System > Hardware > Controller Enclosure.
b. Select the appropriate controller (Controller 1 or Controller 2) to halt.
c. Click the Shut down tab.
d. In the Halt Controller section of the Shut Down Controller window, click Halt. This halts controller processing while the power remains on.

[damiri]

is this already available latest firmware (09500000)?

[James Ackerly]

Yes, I did find the controller replacement guide after my post last night so am happy with that process.

If your master controller goes down does this role automatically switch to the second controller or do you lose access to commandview during this stage?

[damiri]

goes automatically

[damiri]

new firmware

http://h20392.www2.hp.com/portal/swdepot/displayProductInfo.do?productNumber=T4256-63121