I just purchased a ReadyNAS 104 (well, I know it's not designed for business use...). Anyways, it offers iSCSI.
I created an iSCSI LUN- which is seen by my Linux (Centos6/64) box. I can easily access it. So far it works. It is located on a VMware ESXi 5.0U1 as a VM. The ESX host itself it not very busy. The switch is a HP ProCurve 1810G - 24 GE.
As soon as there is (mid to heavy) I/O on the LUN I'm getting connection errors on my Linux host 😞
connection1:0: ping timeout of 5 secs expired, recv timeout 5, last rx 4295124378, last ping 4295129378, now 4295134378 connection1:0: detected conn error (1011)
This error message appears more or less frequently- but as soon as I see it the performance drops! Down to ~50K/sec (even less!) for some minutes...
What I've done for troubleshooting so far:
Google suggests to change some timeout settings so it won't ping (which doesn't affect the issue itself- didn't try) Disabled the tcp_window_size_scaling as suggested by IBM: http://www-01.ibm.com/support/docview.wss?uid=swg21420774 (reduced the frequency but didn't fix it) Tried different vNICs (VMXNET2, VMXNET3, E1000) Enabled Jumbo Frames (everything failed even though it's supported by all devices- strange!) Did a Wireshark trace (which didn't reveal any issues for me except Wireshark didn't capture some packets, but captured the ACK)
So is there anyone who has at least a single clue what is going on here? Why do I get these messages? Why drops the performance then?
The 104 is very low powered low performance ARM CPU. You may just be pushing the system to its limits. The RN304 being x86 would have more power to keep up.
