My setup is 2 x 1GE Bonded to serve the clients and 2 x 1GE bonded for replication on each iSCSI. The replication NICs connect directly to each other, the NIC that serves the client use 2 switches.
When I create a new failover iSCSI and they are syncing for the first time, everything else slows to a crawl, even client that access a totally seperate RAID array on the other nics.
The network between the client and DSS is not the bottle neck, according to DSS and switch graphs.
The network between the iSCSI is running at around 400mbit / (200mbit on each of the NICs), which is due to the replication sync.
Why would the replication sync of a new array, affect the speed of another array connected through a different network ?
With no replication running I get about 40mbit (which also seems slow for a 2 x GE network, )out of hdparm but if there is a replication syncing for the first time on ANY array at all I get 4Mbit from HDPARM.
The RAID card should have a 2.5Gbit throughput, I just can't work out where the bottle neck would be. I've tried different RAID configs (RAID 1 and 10) but the speeds and behaviour is exectly the same).
So, how can the sync between the 2 nodes slow the whole system - all arrays and through all NICs ?
Can you send in the log file from the GUI in Status > Hardware > logs I want the engineers to see what is happening and they will need to look at the whole logs.