Fujitsu Eternus DX600S3 path issue

book

Article ID: 100040173

calendar_today

Updated On:

Description

Error Message

# vxdmpadm getdmpnode
NAME                 STATE        ENCLR-TYPE   PATHS  ENBL  DSBL  ENCLR-NAME
============================================================================
==
disk_0               ENABLED      Disk         1      1     0     disk
fj_dxm0_0000         DISABLED     FJ_DXM       1      0     1     fj_dxm0
<<<
fj_dxm0_0001         DISABLED     FJ_DXM       1      0     1     fj_dxm0
<<<
fj_dxm0_0002         DISABLED     FJ_DXM       1      0     1     fj_dxm0
<<<
fj_dxm0_0003         DISABLED     FJ_DXM       1      0     1     fj_dxm0
<<<
fj_dxm0_0004         DISABLED     FJ_DXM       1      0     1     fj_dxm0
<<<
fj_dxm0_0005         ENABLED      FJ_DXM       2      1     1     fj_dxm0
fj_dxm0_0006         ENABLED      FJ_DXM       2      1     1     fj_dxm0
fj_dxm0_0007         ENABLED      FJ_DXM       2      1     1     fj_dxm0
fj_dxm0_0008         ENABLED      FJ_DXM       2      1     1     fj_dxm0
fj_dxm0_0009         ENABLED      FJ_DXM       2      1     1     fj_dxm0

Cause

After the controller module is going to maintenance mode, the paths belonging to that controller module become disabled,
For unknown reason of disk array side, there is only one Target Port Group Descriptor inquired from the available path belonging to another CM, e.g. following two paths available for RTPG inquiry

/etc/vx/diag.d/vxalua -d /dev/rdsk/c8t500000E0DA896635d4s2    < ----------- RTPG information -----------

------- Target Port Group Descriptor 1 -------
Target port group : 0x8095
Preferred bit : 0x0
Asymmetric access state : 0x1 (Active/Non-Optimized)
Supported states : 0x7 (S_SUP|AN_SUP|AO_SUP|)
Status Code : 0x2
Vendor Specific : 0x0
Target Port Count : 0x1
Relative Target Ports : 0x4095

/etc/vx/diag.d/vxalua -d /dev/rdsk/c8t500000E0DA896635d5s2  << removed path
----------- RTPG information -----------

------- Target Port Group Descriptor 1 -------
Target port group : 0x8095
Preferred bit : 0x1
Asymmetric access state : 0x0 (Active/Optimized)
Supported states : 0x7 (S_SUP|AN_SUP|AO_SUP|)
Status Code : 0x2
Vendor Specific : 0x0
Target Port Count : 0x1
Relative Target Ports : 0x4095

In this case the 2nd path is removed because of Preferred bit : 0x1, DDL thinks it’s an A/A path and we don’t support two different ATYPE for a single array. Eventually the corresponding dmpnode has only one disabled path so that itself gets disabled too

Resolution

This is not our bug as we had implemented the ALUA/ A/A detection logic was provided by Fujitsu to Veritas. Also, DDL can't change the array type on the fly by design.

We could integrated into our code if Fujitsu provides Veritas with the document about how to handle this scenario. Otherwise, We don't need to any changes from our side. However, We have one applicable workaround below:

The CM mode setting in DS600S3 array
(i) Active - Active CM mode
(ii) Active - Standby CM mode(default) and All LUNs should be assigned to active CM

This customer could avoid the problem once one of CMs is going to enter the maintenance mode if this customer applies one of two ways in the workaround.

Issue/Introduction

Customer is using Fujitsu EternusDX600S3, There are two paths (c6, c8).
In case of the controller module(CM) is going to enter the maintenance mode in the array side, Some of dmpnodes get disabled even though there is one available path.