Unsolved
4 Posts
0
1584
November 19th, 2020 07:00
SCv3020 MPIO
I have a doubt regarding iSCSI MPIO in the SCv3020 in a Server 2019 Hyper-V 2 node Cluster, my configuration is the following:
In the servers I use Dell Storage manager Client option “Configure this host to access a Storage Center” for an automatic host configuration, following the manual “Dell EMC SC Series: Microsoft Multipath I/O Best Practices”.
When I map a LUN in the cluster, running the command mpclaim I see that this LUN has 2 MPIO paths active in each server:
Server 1 – iSCSI1 -> Fault Domain 1 Top Controller
– iSCSI2 -> Fault Domain 2 Top Controller
Server 2 has the same configuration:
Server 2 – iSCSI1 -> Fault Domain 1 Top Controller
– iSCSI2 -> Fault Domain 2 Top Controller
Both servers connect to the same controller.
My question is shouldn’t there be 4 MPIO paths (see ex. bellow) or the SCv3020 can only access a LUN with one controller at a time?
Server 1 – iSCSI1 -> Fault Domain 1 Top Controller
-> Fault Domain 2 Bottom Controller
– iSCSI2 -> Fault Domain 2 Top Controller
-> Fault Domain 2 Bottom Controller
DELL-Sam L
Moderator
•
7.6K Posts
0
November 19th, 2020 17:00
Hello r.sys,
Are you using legacy port mode or virtual port mode?
r.sys
4 Posts
0
November 20th, 2020 06:00
Hi Sam L,
Is in Virtual Port Mode.
DELL-Sam L
Moderator
•
7.6K Posts
0
November 20th, 2020 16:00
Hello r.sys,
What is your current SCOS version?
r.sys
4 Posts
0
November 21st, 2020 03:00
Hi Sam,
It's 7.04.10.027.02
DELL-Sam L
Moderator
•
7.6K Posts
0
November 25th, 2020 14:00
Hello r.sys,
The reason that you are only seeing 2 paths is due to your volumes being own by only one controller. if you were to change one of your volumes to the other controller then you will see all 4 paths.
r.sys
4 Posts
0
November 26th, 2020 01:00
Origin3k
4 Operator
•
2.3K Posts
0
November 26th, 2020 04:00
In the world of Compellent and most* of the other so call Active/Active storage systems a LUN is always assign and owned by one** of the controllers. If you dont assign them manually the systems assigned them in a round robin fashion at first or assign a new volume to the less busy controller because different volumes can have different workloads.
If a host try to access a LUN trough the not owning controller it doesnt mean that the access is blocked instead the the try will re redirected the IO or the controller will inform the host that this is not the optimal path. That where ALUA can come into the game.
* Systems exists which are true active/active
** This is one of the reason why one single large volume for everything is often not a good idea because one controller took the complete load and the other one idle around