PRIMECLUSTER Global File Services Configuration and Administration Guide 4.1 (Linux)

B.1.2 sfcfsck(8) Verify and restore GFS Shared File System consistency

<SYNOPSIS>

sfcfsck [generic_options] [-o specific_options] special

<DESCRIPTION>

The sfcfsck command checks the GFS Shared File System for consistency and restores the file system.

The GFS Shared File System can perform high-speed recovery by using the update log function. Specify whether to perform high-speed recovery by executing an update log replay when sfcfsck is executed. If an update log replay cannot be executed because of a damage or for some other reason, sfcfsck executes a full file system check. A file system to be checked can be specified by setting a block special device or a character special device, or it is determined by the mount point of the file system whose entry is contained in /etc/fstab.

For special, specify a block special device or a character special device where the file system has been configured, or specify the mount point. sfcfsck is not executed when a block special device has been mounted.

If there is a serious inconsistency, the operator is prompted to reply yes or no to repair before sfcfsck executes recovery. If the operator does not have write permission for the file system, sfcfsck executes the operation with the -n option (no corrections are made).

sfcfsck executes automatic recovery for minor inconsistencies such as:

• i-node not referenced
• Too many links in an i-node
• An incorrect block included in a free list
• A block appearing in both a free list and a file
• An invalid value detected in a super block

After the inconsistency is corrected, a message is displayed. When recovery of the file system is successful, sfcfsck displays the number of files on the file system, the counts for used blocks and free blocks, and the fragmentation rate.

A full check is executed for the following inconsistencies:

• A block acquired by multiple i-nodes or free lists.
• A block requested by an i-node or free list outside the range of the file system.
• Invalid link count
• Invalid directory size
• i-node with invalid syntax
• A block not referenced from anywhere
• Directory check, a file pointing to an unallocated i-node, i-node number falling outside the range, the first two entries (. and ..) in a directory
• Super block check: i-node blocks exceeding the number of blocks in the file system.
• A free block list with invalid syntax
• An invalid total for the free block count and free i-node count

Independent files and directories (allocated, but not referenced) are placed in a lost+found directory and reunited with the operator's consent. They are assigned i-node numbers as names. If a lost+found directory does not exist, create it. If area for the lost+found directory is not sufficient, expand the directory.

<OPTIONS>

The generic-options consist of the following options:

-v

Displays a command line, but does not execute the command.

-m

Executes a check, but does not make corrections.

One of the following messages is outputted at the time of execution.

sfcfsck: sanity check: device okay

device is consistent.

sfcfsck: sanity check: device need log replay

device needs to perform update log replay.

sfcfsck: sanity check: device needs checking

device needs to perform full check of file system.

Update log replay is automatically performed if needed at the time of mount.

The performing method of the update log replay or full check by sfcfsck is described in -o option.

-n|N

Automatically replies "no" to all inquiries. Does not open the file system for writing.

-y|Y

Automatically replies yes to all inquiries.

-o specific_options

Allows options specific to the GFS Shared File System to be specified. Insert a comma delimiter between options.

f

Executes a forcible check, regardless of the super block status flag.

p

Checks for non-conversational synopsis and makes corrections.

w

Checks only for write-enabled file systems.

log|elog|nolog

Specifies whether to perform an update log replay.

The default is log.

log

Executes an update log replay. Executes a full file system check if the log data is physically destroyed.

elog

Executes an update log replay. The command terminates immediately if the log data is physically destroyed without a full file system check being executed.

nolog

Executes a full file system check without executing an update log replay.

<OPERAND>

The following operand can be specified:

special

Specifies the path to a block special device or a character special device on the representative partition of the GFS Shared File System or the path to the mount point.

<EXAMPLES>

# sfcfsck /dev/sfdsk/gfs/dsk/volume1
# sfcfsck -o nolog /dev/sfdsk/gfs/dsk/volume1

<EXIT STATUS>

0 No recovery was necessary.

1 File system recovery has been completed normally.

2 The system must be rebooted.

4 The file system error has left unrepaired.

8 The processing has terminated because of an abnormal event occurring during command execution.

16 An incorrect parameter was specified.

<SEE ALSO>

sfcmkfs(8).

reboot(8), stat(2), and fstab(5) in the online manual.

<NOTE>

• A file system can be checked only when all hosts that share the file system have started operation. The reason is that the information for these hosts is required.