02-20-2004 05:54 AM
I experimented the same. Not only FC traffic was in the DWDM, but also network traffic, and the public and private networks for clusters were in that DWDM, so with the failover, we had lots of problems...Our customer didn't want to pay for an extra DWDM, so we tried to localize traffic on our 2 fabrics as much as possible to avoid I/O errors during DWDM failovers. We have only backup I/O and data replication in the DWDM, and now we have no more problems.
04-08-2004 02:58 PM
Depending on your DWDM kit , you may or may not see such behaviour . Most DWDM kit have protected lines and will flip over pretty quick . Sometimes the timing is enough to cause the fibre link to detect LOS and your ISL goes down , sometimes it does'nt ....Best Fix as above is to have redundany - dual stretched fabrics , dual fabric connections per stretched fabric each using a different DWDM infrastructure , using different vendors dark fiber ... then you can take a hit on one link and not all traffic in the fabric gets affected although if you are unlucky you may end up losing the path to the principle switch and see a little more activity/timeouts as the fabric goes through principle switch selection again !