.\" .PH "''''"
.\" .PF "''\\\\nP''"
.\" .ds HF 3 3 3 3 3 3 3	\" all headings are bold
.\" .nr Hb 7
.\" .nr Hs 7
.\" .pl 66
.\" .ll 70
.\" .po -5
.\" .ad b
.H 2 "NFS/Diskless Tests (Jason Zions)"
.sp 2
.nf
RESPONSIBLE ENGINEER: Jason Zions
DEPENDENCY:
ISSUE: Final decision on NFS-hetergeneous clusters
.fi
.sp 2
.H 3 "Introduction"
.sp 1
The objective of the NFS/Diskless tests is to isolate defects
peculiar to NFS and Diskless
interactions which would not be found through other 
testing processes.
.sp 1
The success or failure of the tests will be determined using the following
criteria outlined below and if appropriate, an analysis of the end state
of the system.
.sp 1
.BL
.LI
None of the test systems should crash or hang, including otherwise
uninvolved diskless clients.
.LI
The integrity of both services should be preserved.
.LI
No failures occur other than those seen in non-diskless scenarios.
.LE
.sp 2
.H 3 "Intra-cluster testing"
.sp 1
.nf
IMPLEMENT TIME:  0.0  md 
PERFORM TIME:    4.0  md
.fi
.sp 1
Run the full functional and stress suites on a three node cluster, i.e.
.DS
    =============================================
    |              |             |              |
---------      ---------     ---------      ---------
|  9K   |      |  9K   |     |  9K   |      |  9K   |
| Server|<-----| client|     | client|      | client|
---------      ---------     ---------      ---------
|  disk |<---------------------/  Single-line indicates
---------                  NFS client & server relationship
  |__^
.DE
.sp 1
Two of the three clients, as well as the server, should actively participate
in the tests. The third client should run a simple set of non-NFS tests
which merely ensure the node remains busy and that any failure is immediately
visible. One possibility for this test set is the DUX kernel validation
suite, available from SSO.
.sp 1
The goal for this scenario is
48 hours continuous operation at level 2 stress.
Steps should be taken to
ensure that moderate to high filesystem and swapping activity is
maintained during the test.
.sp 2
.H 3 "Inter-cluster tests"
.sp 1
.nf
IMPLEMENT TIME:  0.0  md 
PERFORM TIME:    8.0  md
.fi
.sp 1
Set up five systems in the following configuration:
.sp 1
.DS
    ==============================
    |              |             |
---------      ---------     ---------
|  9K   |      |  9K   |     |  9K   |
| Server|<-----| client|     | client|
---------      ---------     ---------
|  disk |<---     /
---------    \e   /  Single-line indicates NFS
  ^   |       \e /   client to server relationship
  |   |        X
  |   v       / \e   Double-line connects cnodes
---------    /   \e  in the same cluster
|  9K   |<---     \e
| Server|      ---------
---------      |  9K   |
|  disk |      | client|
---------      ---------
    |              |
    ================
.DE
.sp 1
As in the intra-system test scenario, four of the five systems
run functional tests as indicated in the diagram. The fifth
system, a diskless client, does not run NFS tests; it simply
keeps fairly busy performing normal filesystem activities.
.sp 1
Since this configuration is fairly representative of typical
usage, the tests should be run for 120 hours at level 3 stress.
Steps should be taken to
ensure that moderate to high filesystem and swapping activity is
maintained during the test.
.sp 2
.H 3 "Extra-cluster tests"
.sp 1
.nf
IMPLEMENT TIME:  0.0  md 
PERFORM TIME:    6.0  md
.fi
.sp 1
Set up five systems in the following configuration:
.sp 1
.DS
    ===========================================
    |              |             |            |
---------      ---------     ---------    ---------
|  9K   |      |  9K   |     |  9K   |    |  9K   |
| Server|<-----| client|     | client|    | client|
---------      ---------     ---------    ---------
|  disk |                        /
---------        _______________/
  ^   |         /
  |   |        /
  |   v       /  Single-line indicates NFS
---------    /   client to server relationship
|  9K   |<---
| Server|        Double-line connects cnodes
---------        in the same cluster
|  disk |
---------
.DE
.sp 1
As in the above scenarios, four of the five systems
run functional tests as indicated in the diagram. The fifth
system, a diskless client, does not run NFS tests; it simply
keeps fairly busy performing normal filesystem activities.
.sp 1
Since this configuration is fairly representative of typical
usage, the tests should be run for 120 hours at level 3 stress.
Steps should be taken to
ensure that moderate to high filesystem and swapping activity is
maintained during the test.
.sp 2
.H 3 "Error testing"
.sp 1
.nf
IMPLEMENT TIME:  0.0  md 
PERFORM TIME:    2.0  md
.fi
.sp 1
Configure a three system cluster as shown below.
.DS
    ==============================
    |              |             |
---------      ---------     ---------
|9K   A |      |9K   B |     |9K   C |
|Server |      |client |     |client |
---------      ---------     ---------
| w/NFS |      |w/o NFS|     | w/NFS |
---------      ---------     ---------
.DE
.sp 1
.nf
test #     Description                  Response

  1        Bring up A and C; attempt	B should panic during boot
	   to bring up B.               with an appropriate error

  2        Bring up A and C; mount a    umount should fail with
	   remote disk from C; cd to    "directory busy" error
	   the mounted tree on C;
	   attempt to umount on A.

  3        Same as case 2, reversing    umount should fail with
	   roles of A and C.            "directory busy" error
.fi
