05-07-2012 02:36 PM
I've been working on a Bachelor's diploma and would be thankful for some information about Fabric OS.
Which method of data access is used in FOS?
And what does FOS do in case of physical transaction interruption, for example power outage after data have been written to disc but corresponding pointer has not been written yet?
05-08-2012 10:43 AM
--->>> And what does FOS do in case of physical transaction interruption, for example power outage after data have been written to disc but corresponding pointer has not been written yet?
Nothing. The FOS = Fabric Operating System, is a part of Firmware wich allow to make run Brocade Plattforms.
FOS cannot avoid in a case of power outage/interruption to continued to write to Disk after power is recovered.
However, Brocade SAS Application which is used in FA4-18 and Brocade 7600 is ( restricted ! ) capable to continued to write to Disk in in a case of Transfer interruption.
05-15-2012 12:10 PM
Well, I understand that FOS can't continue writing to disc.
But after power outage there must be an autobackout. Maybe somewhere I can find information about it?
And how buffer flushing is organized - is a mystery for me too
05-16-2012 01:47 AM
Do you really got the point that it's just the firmware of the switches and not an OS on the hosts actually doing I/O against storage subsystems?
If you are aware of that and really write a diploma thesis about the different switch firmwares in the market:
FabricOS is based on Linux and uses the normal Linux block I/O services to read and write data to/from its internal non-volatile flash memory. Beside of a firmware update, generating a data collection (supportsave) or writing a dump in case of a panic these data transactions are very short. To avoid data loss in case of parts of the flash being corrupted or I/O disrupted due to power outage the data is written on a primary and a secondary partition of the flash. So if the data is corrupted on one partition (and it's not recoverable with the filesystem tools available as a root user) you could at least boot the switch on the second partition and try to repair it via a firmwaredownload (or at least save your configuration before you replace the switch). The safest possibility is to use a SAN director. It has 2 CP cards and therefore 4 usable partitions in total. So in case of a complete failure of the flash on one CP card you could further run the SAN I/O and let the second CP manage the switch while the faulty one is replaced.
I have to admit that these cases of data corruption are very very seldom.