Friday, July 15, 2011

OCFS2 Guide Lines for 10gR2 CRS and RAC on Linux x86_64

If you plan to install Oracle RAC on Linux using OCFS2 for CRS files and RAC DB files the follow the recommendations for the parameter mentioned , Recommendations are for Version 1.4.7

When Creating the OCFS2 Cluster


It should be set to the timeout value of the io layer. Most multipath solutions have a timeout ranging from 60 secs to 120 secs.
For 60 secs, set it to 31. For 120 secs, set it to 61 according to the follwoing formula : O2CB_HEARTBEAT_THRESHOLD = (((timeout in secs) / 2) + 1)
In general a value of 61 is suitable in most cases

Change the value in the vi /etc/sysconfig/o2cb on all nodes of the OCFS2 Cluster and restart the cluster on all nodes.

When formatting an OCFS2 filesystem.

Cluster Size:

The cluster size is the smallest unit of space allocated to a file to hold  file data.
Permissible cluster sizes include; 4, 8, 16, 32, 64, 128, 256, 512, and 1024Kb.
Oracle recommends a cluster size of 128 KB or higher for datafiles and a cluster size of 32 or 64Kb for Oracle Home volumes or for OCRF and CSS Files.

ATTENTION : The cluster size cannot be modified after OCFS2 volume format.

Block Size:

The smallest unit of space addressable by the filesystem.
Permissible block sizes include; 512 bytes (not recommended), 1Kb, 2Kb, or 4Kb (recommended for most volumes).
4Kb Block Size works fine and Oracle Database Block should be equal or multiple to this 8K , 16K etc.

ATTENTION : The block size cannot be modified after OCFS2 volumes format.

When mounting an OCFS2 filesystem


Use this OCFS2 mount option to mount volumes storing Oracle datafiles, control files, redologs, archivelogs, Oracle clusterware files (voting disk and cluster registry), etc.
The datavolume mount option ensures the Oracle database processes open datafiles with the o_direct flag, required by Oracle Real Application Clusters (RAC) to concurrently access shared files.


Users should add the "noatime" mount option to the mounts that hold the Oracle datafiles, redologs, archivelogs, voting file, etc. This is because OCFS2 1.4 supports atime (access time) which makes it update inodes even on a read. Like ls -l. As tracking access time is not useful for the database, users are advised to disable it for such mounts.

References :
OCFS2 1.4.7-1 Release Notes (Metalink Note ID 1086231.1)
Meatalink Note ID : 727866.1

No comments:

Post a Comment