Out of Sync: Maddy's Blog

Standalone QFS 4.6 FCS and later is supported as a failover filesystem in Solaris Containers managed by SC HA agent for Solaris Container 
with Solaris Cluster 3.2 2/08 and later.

This configuration is supported with Solaris 10 5/08 and later, on SPARC and x64, with HA Solaris Containers and HAStoragePlus. 

Zen-Nippon Shokuhin Co., Ltd , which administers Zen Nishoku Chain of super markets in Japan has selected Solaris Cluster to support an Oracle 10g database with Real Application Clusters running on E6900 systems and Solaris 10 OS. The customer has moved from a main frame based solution to one based on less expensive open architecture.  For more details, you can visit  http://www.sun.com/customers/index.xml?c=zen_nippon.xml.

The combination of Solaris 10 OS, Solaris Cluster and StorageTek QFS provides best of the breed performance and reliability for Oracle RAC deployments and is a good alternative to migrate from existing proprietary hardware/OS at reduced cost.

Equally important is that all the produces are open sourced!  For details on the open source release of the Solaris Cluster software, visit the Open HA Cluster  community at opensolaris.org.

Sun StorageTek QFS file system is a very popular file system for HPC, highly scalable applications due to its good performance and archiving facilities. But somehow, I never tried it out and managed to be ignorant for 2 years. So when the folks asked me to take ownership of a new, very popular project (can't reveal the secret yet but it sure rocks !) it gave me the heebe jeebies. Here I am learning  forced to learn it, with little fear and a lot of excitement.

Shared QFS is mostly used for Oracle RAC in Sun Cluster and it is a very popular one too. I finally managed to configure a very basic setup with the help of a colleague from MDE.

Installation requirements:

1.) Need the SUNWqfsu and SUNWqfsr packages.

 2.) Shared disks between the cluster nodes.

Configuration:

 1.) Copy the mcf, samfs.cmd, hosts.<fsname>  files from the example directory to the /etc/opt/SUNWsamfs directory.
 cp -i /opt/SUNWsamfs/examples/mcf /etc/opt/SUNWsamfs/

2.) For the most basic configuration, we need 1 slice for meta data and 1 for user data.  Also use the cluster interconnects for the meta data transfer between nodes and mention the "shared" key word to denote shared qfs file system. The following operation needs to be done on all cluster nodes:

bash-3.00# cat /etc/opt/SUNWsamfs/mcf
# For additional information about the format of the mcf file, type "man mcf".
#
# This is the mcf file which was built up line-by-line in the
# "Configuring the Sun QFS File System" step in the
# "Sun QFS Initial Installation Procedure" chapter in the
# "Sun QFS, Sun SAM-FS, and Sun SAM-QFS Installation and
# Configuration Guide".  You can read that step
# for more information about this particular mcf.

# Disk cache configuration for 2 file systems: qfs1, qfs2
#
# Equipment         Eq   Eq     Family   Device   Additional
# Identifier        Ord  Type    Set     State    Parameters
#-----------        ---  ----   ------   ------   ----------
#
qfs1                 10    ma   qfs1       on   shared
/dev/did/dsk/d5s0    11    mm   qfs1       on
/dev/did/dsk/d5s6    12    mr   qfs1       on

**It is recommended to use the global names for the shared disks to ensure consistency. EQ type details:

ma - fsname
mm - metadevice
mr - data device

3.) Create the samfs.cmd file with required properties on all cluster nodes.

bash-3.00# cat /etc/opt/SUNWsamfs/samfs.cmd
fs = qfs1
        stripe = 1
        sync_meta = 1
        mh_write
        qwrite
        notrace
        nstreams = 1024
        rdlease = 300
        aplease = 300
        wrlease = 300

 4.)  Create the hosts.<fsname> file on all cluster nodes

bash-3.00# cat /etc/opt/SUNWsamfs/hosts.qfs1
#
# shsam1 basic config, titan/tethys servers, mimas/dione clients
#
# This file goes in titan:/etc/opt/SUNWsamfs/hosts.shsam1, and
# is used by mkfs -S to initialize the FS meta data.  Subsequent
# changes to the configuration are made using samsharefs(1M).
#
# titan/tethys also each have a private interface that they use
# to communicate with each other.
#
ppalomar1       clusternode1-priv       1 - server
ppalomar2       clusternode2-priv       2 -

5.) On all cluster nodes, set the environment variables :

 bash-3.00# cat .bashrc
export PATH=$PATH:/usr/cluster/bin:/usr/ccs/bin:/usr/sfw/bin:/opt/SUNWsamfs/bin:/opt/SUNWsamfs/sbin:/opt/SUNWsamfs/tools
export MANPATH=/usr/man:/usr/cluster/man:/opt/SUNWsamfs/man:/opt/SUNWsan/man

6.) Start the QFS daemons on all the cluster nodes:

# samd config

7.) Create the filsystem:

 # sammkfs -S <fsname>

8.) Make the entries in /etc/vfstab on all the nodes:

9.) Mount the filesystem!

10.) Register the QFS Resource type and add it to the RG!

#scrgadm -at SUNW.QFS

#scrgadm -at SUNW.qfs -j qfs -g qfs-rg  -x QFSFileSystem=/global/fs

It is as simple as that!  

Now that makes me wonder why I had the heebe jeebies on the first place!

P.S: Since  I am working on a lot of hot shot projects, I am thinking of some black magic to ward off evil eyes.