freezefs - The freezefs() API causes the specified fileset's
domain to enter into a metadata stable state, causes
a previously frozen fileset's domain to resume normal
activity, or queries the freeze state of the fileset's
domain. All the filesets in the AdvFS domain are affected.
#include <sys/mount.h>
int freezefs {
char *mountPoint,
u_long optionFlags,
int timeout };
);
Points to a null-terminated string that contains the
appropriate mount point of the file system. Specifies
what action to take on the specified file system.
FS_Q_QUERY
Queries the freeze state and indicates whether or
not the domain is frozen by way of the return code.
FS_Q_FREEZE
Places the filesystem in a metadata consistent
state and guarantees that it stays that way until
thawed. All metadata, which could be spread across
multiple volumes or logical units (LUNS), is
flushed to disk and does not change for the duration
of the freeze.
FS_Q_THAW
Causes a previously frozen fileset's domain to
resume normal activity.
timeout
Specifies the maximum time allowed for the fileset's
domain to remain frozen:
> 0
Specifies the timeout value in seconds.
= 0
Uses the default timeout specified in the vfs
stanza of /etc/sysconfigtab as
freezefs_default_timeout. If not specified in
/etc/sysconfigtab, freezefs uses 60 seconds.
< 0
Specifies no timeout. The fileset's domain remains
frozen until explicitly thawed.
Certain system events on a cluster will cause automatic
thaws regardless of timeout value. See DESCRIPTION for
more information.
To allow coherent hardware snapshots in multivolume domain
configurations, filesystem metadata must be consistent
across all volumes when the individual volumes are snapped
or cloned. These configurations include both multivolume
AdvFS domains and multivolume LSM logical volumes.
The FS_Q_FREEZE option to the freezefs API places the
filesystem in a metadata-consistent state and guarantees
that it stays that way until thawed. All metadata, which
could be spread across multiple volumes or logical units
(LUNs), is flushed to disk and does not change for the
duration of the freeze.
The filesystem thaws either by timing out or explicitly
with the FS_Q_THAW option to the freezefs() API. If you
are running a cluster configuration, shutting down any
node or the failure of any node will also thaw the
filesystem.
The exact nature of the freezefs API depends on whether
you are running on a single system or a cluster.
Single System
When you freeze a filesystem, any in-process metadata
updates are allowed to finish. Filesystem operations that
do not modify metadata are allowed while frozen. Some
operations will work normally even if the target fileset's
domain is frozen, for example: Read No-extending write
Stat
Filesystem operations that modify metadata are blocked.
The following operations will block if the target fileset's
domain is frozen, for example: Mount of fileset
rmvol Extending write mkfset / rmfset
The following filesystem operations will fail immediately
if the target filesystem (that is, domain) is frozen:
Mount update of fileset Unmount of fileset
Cluster
When you freeze a filesystem in a clustered configuration,
all in-process filesystem operations are allowed to complete.
Some filesystem operations are allowed while
frozen. The following operations, which do not require
metadata updates, will work normally even if the target
filesystem is frozen. Read Stat
Most new filesystem operations are blocked when the
filesystem is frozen. Operations that have the potential
for requiring metadata updates will block if the target
filesystem is frozen, for example: Write chmod Link
Some filesystem operations will fail immediately if the
target domain is frozen, for example: Unmount of fileset
Mount (update) of fileset User-initiated planned relocation
User-initiated forced unmount addvol rmvol
The freezefs() API posts an event manager (EVM) event when
a file system is frozen or thawed. Use the evmwatch and
evmshow commands to determine if any file systems in the
cluster are frozen. See EXAMPLES.
At present, freezefs and thawfs functionality is only
implemented for AdvFS filesystems.
Using the freezefs() API for a single mount point affects
all filesets in the domain.
In a cluster configuration, some system events may cause
an automatic thaw regardless of the timeout option. See
the DESCRIPTION section for more information.
You must be root user to run freezefs and thawfs. The /,
/usr and /var file systems cannot be frozen.
Success FS_Q_FREEZE and FS_Q_THAW
Not frozen FS_Q_QUERY Frozen FS_Q_QUERY An error
occurred.
The freezefs is not supported by the filesystem specified.
The filesystem is not mounted. Not enough memory. The
file system is already frozen freezefs, or
The file system is not frozen thawfs. The caller
does not have root privilege. Option not supported.
To freeze a file system: #include
extern errno;
...
char *mp;
u_long flags;
int timeout, rc;
...
mp = "/oracle"; /* filesystem to freeze */
flags = FS_Q_FREEZE; /* freeze flag */
timeout = 0; /* use default timeout */
...
errno = 0;
rc = freezefs(mp, flags, timeout);
if (rc = -1) {
error
}
.
.
.
PERFORM WORK ON FROZEN FILE SYSTEM
.
.
.
flags = FS_Q_THAW; /* thaw flag */
errno = 0;
rc = freezefs(mp, flags, timeout);
if (rc = -1) { /* If error on thaw, cannot
rely on the intial freeze */
error
}
Commands: freezefs(8), thawfs(8)
freezefs(2)
[ Back ] | 