This is mentioned in the doc/FAQ. Ocfs2 does not update the mtime on disk for
non-extending directio writes. The correct times you are seeing are only in
cache. We do this to allow multiple nodes to r/w concurrently to the same file.
Having said that, in mainline we have recently added a mount option that allows
users to choose the other behavior. It will be made available in some later
release.
On Jan 19, 2011, at 6:09 PM, Jason Price <japrice at gmail.com> wrote:
> # cat /etc/enterprise-release
> Enterprise Linux Enterprise Linux AS release 4 (October Update 5)
>
> # rpm -qa | grep ocfs
> ocfs2-tools-devel-1.2.4-1
> ocfs2console-1.2.4-1
> ocfs2-2.6.9-55.0.0.0.2.ELsmp-1.2.5-1
> ocfs2-tools-1.2.4-1
> ocfs2-2.6.9-55.0.0.0.2.EL-1.2.5-1
> ocfs2-2.6.9-55.0.0.0.2.ELlargesmp-1.2.5-1
>
> The mount point is vrp-db. Two nodes mount the SAN volume. An oracle
> DB runs on one node at a time (no RAC). Currently oracle is running
> correctly on node2.
>
> on node1 (bad modtimes:)
>
> # ls -l /vrp-db/u02/oradata/VRP/
> total 29679423
> -rw-r----- 1 oracle oinstall 14663680 Aug 29 2008 control01.ctl
> -rw-r----- 1 oracle oinstall 14663680 Aug 29 2008 control02.ctl
> -rw-r----- 1 oracle oinstall 14663680 Aug 29 2008 control03.ctl
> -rw-r----- 1 oracle oinstall 1073750016 Dec 14 16:21 lobdata.dbf
> -rw-r----- 1 oracle oinstall 268443648 Dec 14 16:22 lobindex.dbf
> drwxr-xr-x 2 oracle oinstall 4096 Feb 20 2008 ocrmirror
> -rw-r----- 1 oracle oinstall 21474844672 Dec 29 20:56 pilot_data01.dbf
> -rw-r----- 1 oracle oinstall 2147491840 Feb 25 2008 pilot_idx01.dbf
> -rw-r----- 1 oracle oinstall 104858112 Feb 20 2008 redo0101.log
> -rw-r----- 1 oracle oinstall 104858112 Feb 20 2008 redo0102.log
> ...
>
> on node2 (correct modtimes:)
>
> # ls -l /vrp-db/u02/oradata/VRP/
> total 29679423
> -rw-r----- 1 oracle oinstall 14663680 Jan 20 02:05 control01.ctl
> -rw-r----- 1 oracle oinstall 14663680 Jan 20 02:05 control02.ctl
> -rw-r----- 1 oracle oinstall 14663680 Jan 20 02:05 control03.ctl
> -rw-r----- 1 oracle oinstall 1073750016 Jan 20 02:04 lobdata.dbf
> -rw-r----- 1 oracle oinstall 268443648 Jan 19 23:29 lobindex.dbf
> drwxr-xr-x 2 oracle oinstall 4096 Feb 20 2008 ocrmirror
> -rw-r----- 1 oracle oinstall 21474844672 Jan 20 02:04 pilot_data01.dbf
> -rw-r----- 1 oracle oinstall 2147491840 Jan 20 02:03 pilot_idx01.dbf
> -rw-r----- 1 oracle oinstall 104858112 Jan 20 02:05 redo0101.log
> -rw-r----- 1 oracle oinstall 104858112 Jan 20 02:05 redo0102.log
> ...
>
> Both nodes were rebooted this afternoon, so these are fresh mount points.
> What on earth is going on here?
>
> (making things worse... before the reboot, node1 had mod times around
> Dec 10 2010... which is WAY out of wack, but still better than Feb 20
> 2008!)
>
> --Jason
>
> _______________________________________________
> Ocfs2-users mailing list
> Ocfs2-users at oss.oracle.com
> http://oss.oracle.com/mailman/listinfo/ocfs2-users