-3

I'm just wondering whether it's possible for a UNC path to a SAN be disconnected through overwhelming I/O?

If this is possible, how would one recover from it?

This would be a Windows server environment. I'm supposing that the specifics of the SAN are irrelevant.

If further information is required, please let me know.

Thank you.

chuacw
  • 145
  • 7
  • 1
    Please rephrase the question, describing your specific environment and issue. What sort of IO level? what SAN? what do you mean by 'UNC path disconnected'?, etc – BlueCompute Aug 24 '15 at 09:03
  • Why would the SAN be irrelevant???? – BlueCompute Aug 24 '15 at 09:04
  • @BlueCompute If the SAN supports a total of 10K IOPS, and I have 6 separate Windows servers throwing a combined total of 15K IOPS? If the SAN supports a total of 20K IOPS, I have 7 separate Windows servers throwing a combined total of 25K IOPS, etc? – chuacw Aug 24 '15 at 09:46
  • Review the advice here: http://serverfault.com/help/how-to-ask It's very difficult to help without details of the problem encountered... – BlueCompute Aug 24 '15 at 11:50
  • UNC paths are for network drives. SAN provides SCSI LUNs. Unless you clarify a specific scenario that you are dealing with, this question should be closed. – Basil Aug 24 '15 at 12:35

2 Answers2

1

Theoretically yes, but we need to know your environment specifics. What SAN?, whats the presentation (FC, iSCSI, NFS...)? Are you using the SAN to back a file server and are exporting shares from that?

'UNC' paths refer to a directory like structure such as "\host\share\directory1\directory2" which would suggest you are presenting file level shares to your servers instead of direct block such as FC or iSCSI.

Please give specifics.

tomstephens89
  • 1,011
  • 1
  • 12
  • 24
0

Theoretically yes, it is possible for a (really) overwhelming I/O pattern to apparently put the SAN-exported disk off-line.

The point is that if I/O requests take too much time, the remote OS can detect the disks as failed/disconnected, even if they are not really failed or disconnected.

For this very reason, typically a deadline (timeout) is assigned to any I/O request: if the I/O completition take too much time, the request is aborted and marked as failed to the remote OS, but the virtual disk itself remain online. This give the remote OS a chance to re-issue the failed request, without putting the vdisk offline.

shodanshok
  • 47,711
  • 7
  • 111
  • 180