Wednesday Apr 08, 2009

nested mounts may fail to mount in the correct order on Sun Cluster 3.2

In case of Sun Cluster 3.2 it's possible that nested mounts will be mounted in the wrong order. As a result, the data on these file systems become inaccessible to users.

The issue happen if one of the following Sun Cluster core patches are active and nested mounts are managed with resource type SUNW.HAStoragePlus.
126106-27 or -29 or -30 Sun Cluster 3.2: CORE patch for Solaris 10
126107-28 or -30 or -31 Sun Cluster 3.2: CORE patch for Solaris 10_x86
126107-26 or -28 or -29 Sun Cluster 3.2: CORE patch for Solaris 9

The error can look like:
The correct output of df -k should be
/dev/vx/dsk/datadg/vol01 480751 1048 431628 1% /test
/dev/vx/dsk/datadg/vol02 288639 1042 258734 1% /test/test2
/dev/vx/dsk/datadg/vol03 577295 1041 518525 1% /test/test3

The mount order is defined in the HAStoragePlus resource test-rs
# clrs show -v test-rs | grep FilesystemMountPoints
FilesystemMountPoints: /test /test/test2 /test/test3

But, due to runtime problems the filesystems get mounted in wrong order and the df -k can look like:
/dev/vx/dsk/datadg/vol02 480751 1048 431628 1% /test/test2
/dev/vx/dsk/datadg/vol03 480751 1048 431628 1% /test/test3
/dev/vx/dsk/datadg/vol01 480751 1048 431628 1% /test
In this specific case, /test/test2 and /test/test3 were mounted first followed by an overlay mount of /test. Due to this, data in /test/test2 and /test/test3 would not be accessible and show the same information as /test.

Workaround:
It's possible to split the SUNW.HAStoragePlus resource. For the example above change the resource test-rs and remove the FilesystemMountPoints /test/test2 and /test/test3. Furthermore create a new resource test1-rs with the mentioned FilesystemMountPoints and add a resource dependency.
The commands to change this specific configuration will be:
# clrs set -p FilesystemMountPoints=/test test-rs
# clrs create -g test-rg -t SUNW.HAStoragePlus -p FilesystemMountPoints=/test/test2,/test/test3 -p Resource_dependencies=test-rs -p AffinityOn=True test1-rs

Due to this change the test1-rs starts after the test-rs and the problem is solved.
Details available in:
Alert 1020328.1 Nested Mounts Managed by a SUNW.HAStoragePlus Resource may Fail to Mount in the Correct Order on Solaris Cluster 3.2

Update 17.Jun.2009:
The -33 revision of the Sun Cluster core patch is the first released version which fix this issue.
126106-33 Sun Cluster 3.2: CORE patch for Solaris 10
126107-33 Sun Cluster 3.2: CORE patch for Solaris 10_x86

About

I'm still mostly blogging around Solaris Cluster and support. Independently if for Sun Microsystems or Oracle. :-)

Search

Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today