Consider we are having Bus 2 with 5 Enclosures connected to it daisy chained as follows:
What happens when any of the DAE fails completely? Say if DAE 1 is failed, will the connectivity to DAE above it(2,3,4) will fail? Also is there any connection from Enclosure 4 to Bus 2 directly? Or the data from E4 has to traverse from 4 to 0 before reaching Bus 2?
You have 2 LCCs on each bus. LCC = Link Control Card
One LCC for SPA
One LCC for SPB
If on the same bus, enclosure 1 fail (LCC Card, cable...), Enclosure 2 and higher will be unaccess.
All luns in enclosure 1 to 4 tresspass to other SP and use the survival LCC.
So in your design, i think two controllers (not single SP),
if you loose enclosure 1 on left side, Luns affected (in enclosure 1 to 4) will be tresspass to other SP to access it by the right side LCC.
To successfull access to a lun tresspassed, you must have a failover software in host side.
The data from Enclosure 4 will be served by the same SP but the data from failed LCC in case (you need to check whether LCC is failed or Cable or fault in DAE), the Peer LCC for the same enclosure will serve the data, but latency will be a bit higher. SP will work the same way, as it was before.
Thanks Rakesh and Fredh!
What happens when the enclosure1 got powered off because of corresponding pdu(s) failure? I know these are rarest cases. Trying to understand the basics. Will the power failure of enclosure 1 cause it to be go to pass through mode and keep the loop complete? or all the enclosures above DAE 1 will became unavailable?
In this case, you have other pdu, to serve the electricity to your DAE. It only supplies the power to the DAE and doesn't have to do anything with the loop. The only thing will happen that the load on the socket of second pdu will be increased.
Evey part of Storage system is redundant, even if has a slight meaning to the system, however EMC or any manufacturer always considers the worst case like us.
Now if we think that if both redundant devices (any part of system) get failed before replacing the first failed device, which first doesn't happen, and if it happens, array assumes that the peer device is about to be dead, it comes in degraded mode, slows the performance, throws the Alarms to Vendor (EMC) and Client supported teams via E-mail.
Still if Storage Engineers or Vendors could not do anything, then there's a big question on existence of such vendors being still in the Industry & survival of Engineers.