Unsolved
This post is more than 5 years old
4 Posts
0
39882
February 25th, 2013 02:00
equallogic iscsi problem
Hi,
We have an equallogic PS6010 with 16 SAS 16k, set up in Raid50.
If i create a volume and connect it using MPIO with Iscsi with two nics, so we get about 190MB /s from a hyper-v host. (is that ok?).
When creating a fixed disk for a vm on that hyper-v host located on the same volume, we're getting 110MB max when writing to that fixed disk from within the vm.
Any ideas why this is?
Thanks
No Events found!
kjartanius
4 Posts
0
February 25th, 2013 07:00
Yes it is.
kjartanius
4 Posts
0
February 26th, 2013 01:00
The hosts are 10GB.
I already have a support case open on it, nothing usefull yet about this particular problem.
Could it be some queue depth settings or something?
DELL-Joe S
7 Technologist
•
729 Posts
0
February 26th, 2013 06:00
As Don indicated, not easy to troubleshoot over a forum, see if your case can be escalated to the EqualLogic performance team.
Also, a few things you can check on your own:
1. Is Jumbo frame setup correctly on the host, and switch?
2. Ensure that the link between the host, switch and arrays do not have any packet loss. Use ping to test each interface on the host and the array(s), by specifying the interface to ping from and the interface to ping to, you need to do all possible combinations:
To ping from the array:
ping "-I array-ip-address ip-address" (the “array-ip-address” is each array interface the “ip-address” is each of the MPIO interfaces), do all possible combinations. Let ping run for a while (3-5 minutes), use ctrl-c to break out.
From the host back to the array:
Ping -t –s srcaddr destaddr (the “srcaddr” is one of the MPIO interfaces on your host , the “destaddr” is each physical interface on your array(s)). Do all possible combinations; let the ping run for a while, to see if any loss is occurring (like 3-5 minutes or longer).
More than 2% packet loss or high latency (over 20ms) could indicate a problem with your SAN configuration (as in the Storage network, host drivers, switch configuration, switch firmware, etc.).
3. You didn’t mention the switch or NIC you are using, or if you are using DCB.
Not all switches support DCB properly for the arrays. The switch must be configured for Loss less priority or priority flow control (PFC) for iSCSI. This link discusses DCB requirements in detail: en.community.dell.com/.../4396.data-center-bridging-standards-behavioral-requirements-and-configuration-guidelines-by-sis.aspx
4. Also, review the configuration portal site, this site will provide you with the “bare bones” configuration requirements.
en.community.dell.com/.../2632.storage-infrastructure-and-solutions-team-publications.aspx
-joe
David Caughey
2 Posts
0
February 26th, 2013 07:00
Yes the hosts are 10 GIG but your switch architecture is 1 gig, is that not correct??
kjartanius
4 Posts
0
February 26th, 2013 07:00
HI David,
The hosts/switches are 1GB, but the array is 10GB, connected using 10GB modules ont he switch.
But I believe it might be a jumbo frame issue, since when i try to ping with ping -f -l 8792 it's not working correctly. However i've set the jumbo frames, but does it need a reboot? Does the Iscsi only look at the frame when connecting the first time?
Thanks
David Caughey
2 Posts
0
February 26th, 2013 08:00
So where are pinging from to?
No you won't need to reboot just perform a wr to mem.
It may be connected by a 10gig module but you can't shove a 10gig pipe down a pipe that is designed for 1 gig.
Did you adjust the queue depth?
Is the host set to send jumbo frames?
they need to be enabled everywhere.