EVA4400 Command View paths lost

[jpmartin]

Hi,

i have again very hard problems with the command view paths to access storage EVA4400's.

It is working fine only if each EVA controller is restarted.

In a CA configuration, when a storage EVA is power off, all paths of the remaining EVA are lost, shown with the 'cvutil paths' command, and the Command View don't see anything.

Restarting each controller resolve or not resolve this problem.

Question : is there a solution to recover the 4 CV paths of each storage EVA.

Note : each EVA4400 is running 9.522.000, Command View 9.1 and the FC switchs 6.2.1

Thank you for help

Jean-Pierre

[blackhawk_1]

What do the EVA logs have to say ?
Is switch showing any kind if issues on any port ?

[jpmartin]

under inverstigation

[kunalsahoo]

Hi,

this might help your investigation...

CVEVA always talks only to the master controller....when ever master controller reboots the slave becomes Master and that time CVEVA talks to that controller..if all the paths from CVEVA to EVA are fine then there shouldn't be any problem accessing the EVA..

[jpmartin]

Yes, i have seen it, 2 Master paths and 2 Slave paths between EVA and CV.

But the problem is when the 2 EVAs are Power on, the CV would have to see 8 paths to access the 2 EVAs, but some paths disapear, and CV hangs as SSSU.

Today, the 2nd EVA is powered off until i have more elements to continue.

[kunalsahoo]

it has to be a communication issue........

switches also along with multipathing software can be suspected

[kunalsahoo]

it seems to communication error
switches as well as Multipathing software can be suspected

[jpmartin]

I will check next time

[BBouldin]

You might take a look at the output of a program called CVUTIL.EXE that is in the "Element Manager for StorageWorks HSV" folder. If you're using WS2008 be sure to start your command prompt session with administrator privileges. For example,

cvutil paths

will list the SCSI port/target/LUN that CV can see for each path to the controllers. In my configuration I have 2 total paths to the array, one to each controller. One path says the controller is the Master and the other says Slave.

When CV says that it can't manage the array it's because there's no path to the Master controller. This is a handy tool to let you see the paths that CV is using.

Looking at the status of the ports on the switch might help too. I recently had a bad SFP in my switch that caused problems for CV.

Regards,
Brian

[jpmartin]

CVUTIL is the program i used to check the paths, and below when is working fine

C:\Program Files\Hewlett-Packard\Sanworks\Element Manager for StorageWorks HSV>c
vutil paths
Status ................. done.

Path Vendor Product Firmware Node Serial SCMI
Master
WWN Number State
---------------- ------ ------- -------- ------------------- -------------- ----- ------
\\.\Scsi2:-0-1-0 HP HSV300 0952 5001-4380-025B-EA50 P5512G39SWL0K6 Ready Master
\\.\Scsi3:-0-1-0 HP HSV300 0952 5001-4380-025B-EA50 P5512G39SWL0K6 Ready Master
\\.\Scsi2:-0-0-0 HP HSV300 0952 5001-4380-025B-EA50 P5512G39SWL0M3 Ready Slave
\\.\Scsi3:-0-0-0 HP HSV300 0952 5001-4380-025B-EA50 P5512G39SWL0M3 Ready Slave
\\.\Scsi3:-0-2-0 HP HSV300 0952 5001-4380-025C-9920 P5512G39SWP049 Ready Master
\\.\Scsi2:-0-2-0 HP HSV300 0952 5001-4380-025C-9920 P5512G39SWP049 Ready Master
\\.\Scsi3:-0-3-0 HP HSV300 0952 5001-4380-025C-9920 P5512G39SWS0K9 Ready Slave
\\.\Scsi2:-0-3-0 HP HSV300 0952 5001-4380-025C-9920 P5512G39SWS0K9 Ready Slave

and after, when became badly :

C:\Program Files\Hewlett-Packard\Sanworks\Element Manager for StorageWorks HSV>cvutil paths
Status ................. done.

Path Vendor Product Firmware Node Serial SCMI
Master
WWN Number State
---------------- ------ ------- -------- ------------------- -------------- -------------------- -------
\\.\Scsi2:-0-1-0 HP HSV300 0952 5001-4380-025B-EA50 P5512G39SWL0K6 Ready Master
\\.\Scsi3:-0-1-0 HP HSV300 0952 5001-4380-025B-EA50 P5512G39SWL0K6 Ready Master
\\.\Scsi2:-0-0-0 HP HSV300 0952 5001-4380-025B-EA50 P5512G39SWL0M3 Ready Slave
\\.\Scsi3:-0-0-0 HP HSV300 0952 5001-4380-025B-EA50 P5512G39SWL0M3 Ready Slave
\\.\Scsi2:-0-2-0 HP HSV300 0952 5001-4380-025C-9920 P5512G39SWP049 Not valid ---
\\.\Scsi3:-0-2-0 HP HSV300 0952 5001-4380-025C-9920 P5512G39SWP049 Not valid ---
\\.\Scsi2:-0-3-0 HP HSV300 0952 5001-4380-025C-9920 P5512G39SWS0K9 non-responsive retry Unknown
\\.\Scsi3:-0-3-0 HP HSV300 0952 5001-4380-025C-9920 P5512G39SWS0K9 Not valid ---


until no paths found !!!

[Dileep..]

Hello Jean,
Few things which you can check ,

1) Is CV is accessible through WOCP ?
2) Isolate CV server ports from other hosts which access EVA (Single initator). This kind of isues has been observed when we zone the CV server ports with other hosts.

Regards,
Dileep

[jpmartin]

1) Is CV is accessible through WOCP ?

Yes, embedded CV 9.1 on TBM 1.30 is running fine on each EVA4400 to manage its EVA4400, but is not applicable on a CA configuration.


2) Isolate CV server ports from other hosts which access EVA (Single initator). This kind of isues has been observed when we zone the CV server ports with other hosts.

On each Fabric, this is the zoning made for the management zone :

zone: Eva_ManApp1_Fabric1
Eva1_Ctler1_Port1; Eva1_Ctler2_Port1; ManApp1_Port1; Eva2_Ctler1_Port1; Eva2_Ctler2_Port1

zone: Eva_ManApp2_Fabric1
Eva2_Ctler1_Port1; Eva2_Ctler2_Port1; ManApp2_Port1; Eva1_Ctler1_Port1; Eva1_Ctler2_Port1

and each Host I/O zone :

zone: Eva_Esx01_Fabric1
Eva1_Ctler1_Port1; Eva1_Ctler2_Port1; Esx01_Port1; Eva2_Ctler1_Port1; Eva2_Ctler2_Port1

In regard with the new CA Implementation guide, i will keep the management zone, but i will change each Host I/O zone with one initiator for only one EVA, and add a replication zone per Fabric.

Each Host I/O will have two zones per fabric :

zone: Eva1_Esx01_Fabric1
Eva1_Ctler1_Port1; Eva1_Ctler2_Port1; Esx01_Port1

zone: Eva2_Esx01_Fabric1
Eva2_Ctler1_Port1; Eva2_Ctler2_Port1;
Esx01_Port1

and a new replication zone per Faric :

zone: CA_Replication_Fabric1
Eva2_Ctler1_Port1; Eva2_Ctler2_Port1; Eva1_Ctler1_Port1; Eva1_Ctler2_Port1

[Dileep..]

Hello Jean,
Are you saying that in Eva_ManApp1_Fabric1 and Eva_ManApp2_Fabric1 you kept all the four FP ports in same zone ? If so this is not recommended.

Below is the recommendation and this applicable for host as well CV server also.

The best way to zone is (Single initiator and target) , which means

Fabric1 :
CV Server HBA1+ FP1Ctrl1 = zone1
CV Server HBA1+ FP1Ctrl2 = zone2

Fabric2
CV Server HBA2+ FP2Ctrl1 = zone1
CV Server HBA2+ FP2Ctrl2 = zone2

Note: if you have only single switch then separate the ports with zoning is mandatory.If you have single HBA on CV Server then you can ignore zoning with FP2 of both the controller. But its recommended to have dual HBA for redundancy.

I am off for next one week and wont be able to respond you till that. If you need a quick help, then I would suggest to log a case with HP. Other wise I respond you once I am back.


Regards
Dileep

[Dileep.. · Accepted solution]

refer page82 on below link
http://bizsupport2.austin.hp.com/bc/docs/support/SupportManual/c01883137/c01883137.pdf

[jpmartin]

It is confused, i simply used the best practises documented in the CA Implementation Guide page 74 at http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c01800459/c01800459.pdf

Another thread opened yesterday http://forums13.itrc.hp.com/service/forums/questionanswer.do?threadId=1379835&admit=109447627+1256014249579+28353475 talks about a customer advisory for the EVA4400 which requires an even finer granularity at least for the management server:
- HOST1_1,EVA1_C1P1
- HOST1_1,EVA1_C2P1

In a CA configuration, its become for CV Server :

- HOST1_1,EVA1_C1P1,EVA2_C1P1
- HOST2_1,EVA1_C1P1,EVA2_C1P1

and for Hosts I/0 :

- HOST1_1,EVA1_C1P1,EVA1_C2P1
- HOST1_1,EVA2_C2P1,EVA2_C2P1
- HOST1_2,EVA1_C1P2,EVA1_C2P2
- HOST1_2,EVA2_C2P2,EVA2_C2P2

[jpmartin]

Thank you all for tips,

i will setup the new zoning best practises the next time the production is halted.