This post is more than 5 years old
2 Intern
•
227 Posts
0
2331
VMAX 10K Single Engine (Port Group Question)
Still getting used to the symmetrix world and am running into an issue. we have an emc hardware issue that requires a board be replaced. EMC came out today and ran their script to validate HA and found some boot lunes that dont have alternating paths:
Should someone have configured the port group so that is uses a FA port from each director VS two from the same and is that whats causing the script to trigger an issue?
codyhosterman
286 Posts
1
March 26th, 2014 15:00
You typically want to spread as wide as you can when it comes to using front end ports. So if you only have a single engine, for a given host/LUN assign it to a port on both directors. If you have multiple engines you might want to spread across both engines. This is for performance and redundancy. Check out some examples on page 25 and 26 in this techbook for VMware and Symmetrix:
http://www.emc.com/collateral/hardware/solution-overview/h2529-vmware-esx-svr-w-symmetrix-wp-ldv.pdf
Might be using a different host, but the idea remains the same for connectivity.
M_Salem
213 Posts
0
March 26th, 2014 15:00
Duhhas, What is the VMAX model you have and how many engines are installed?
duhaas1
2 Intern
2 Intern
•
227 Posts
0
March 26th, 2014 15:00
thanks a bunch Cody Hosterman so the issue is just that, the only ports I've configured for this host are located on the same director which would cause the issue when part gets replaced?
codyhosterman
286 Posts
0
March 26th, 2014 15:00
Hard to say. Could be any number of reasons I suppose. There is some redundancy protection here in the case of a port failure but not in the case of a director failure. But just to spitball some possibilities it could be that the port settings on the other director are incompatible with whatever host you have connected to director 1 or even incompatible protocols. Maybe the cabling wasnt complete and the hosts couldnt be connected to the other director at the time and it was just never added when the cabling was complete. Maybe the other director is for some reason segregated on another SAN which that host or the switches to that host are not connected to. There are other possibilities but they are increasingly unlikely.
codyhosterman
286 Posts
1
March 26th, 2014 15:00
Sure thing! Yeah most likely. Especially if the part being replaced is the director (1) or the front end slic housing that contains those two ports.
duhaas1
2 Intern
2 Intern
•
227 Posts
0
March 26th, 2014 15:00
thanks again Cody Hosterman any reason why you can think of why it would of been configured this way? i only ask because I know emc was heavily involved with the design and initial configuration and i'm guessing it was at their recommendation that we configured each port group for hosts and clusters to have ports defined from either one side or the other
duhaas1
2 Intern
2 Intern
•
227 Posts
0
March 26th, 2014 15:00
VMaxe
M_Salem
213 Posts
0
March 26th, 2014 16:00
As Cody pointed out many reasons might be behind this design, The best design will be the design that eliminates a SPOF (Single Point of Failure) so always try to mask at least one port from each Dir with same Port FLAGs set and present it to machines running same OS for instance 1f:0 and 2f:0 for Windows hosts, 1e:0 and 2e:0 for Solaris etc..
This configuration should pass the.HA tests. So you should check now for any available FAs ports that can be added to your host port group.For example, For Windows the FLAGs C,OS2007,SPC2,SC3 should be set on the FA port. If you found another FA port with same FLAG set then you can add it to your server port group. The steps to do so should be as follows:
1- Find the eligible FA by running the command: symcfg -sid xx -v -dir all list -->This command will list all FAs you have and the FLAGs set on ports.
2- Zone your host HBAs to the new FA port
3- Add the new FA port to your port group using this command: symaccess -sid xxx -type port -name xxxxx add -dirport fa:port
4- The last command will automatically map your devices to the new FA port so now you can remove one of the FA ports above and to do so
symdev -sid xxx write_disable -fa x -p y
Now, running symdev show against any of your devices should return only 2 FA ports one from each director and this configuration should pass the HA test.
Hope it helps
Mohammed Salem @yankoora
duhaas1
2 Intern
2 Intern
•
227 Posts
0
March 27th, 2014 04:00
thanks to both of M.Salem and Cody Hosterman for the great education and feedback, will help me move past this issue.