Unsolved
This post is more than 5 years old
1 Message
0
521
MDM Degraded error
Hello, we are using 1.35 version of Scale-IO in our test environment, and noticed that once per two hours or under high network loads we have MDM_DATA_DEGRADED error. As I cloud understand, reason for this is networking latency problem, which have to be sub 0,5ms between MDM nodes.
We are using Intel X520 10Gb NICs, with dedicated EMC VDX Switches, with Jumbo frames enabled. All MDM/SDS/SDC traffic goes on same physical network, throe two optical 10Gb links.
9KB ping latency between all four nodes are sub 0,1ms, but when we run query_network_latency_meters among SDS, we have average latency around 500 micro seconds (0,5ms) with average IO size 9KB (9216 Bytes).
Please advise, is it ScaleIO misconfiguration problem, either we have to tune network for it?
Thanks!
pawelw1
306 Posts
0
August 5th, 2016 05:00
Hi Anton,
Can you provide some more information about your environment? Is it virtualized or physical, what kind of OS is that?
Can you see any errors/dropped packets in the interfaces statistics?
Did you follow "Fine-tuning ScaleIO performance" document? If not, it would be worth trying and see if it improves the situation in any way...
Thanks,
Pawel