large partition fsck on shutdown instead of boot

lazardo · 09-15-2010, 02:22 AM

Got tired of long waits for fsck on very large partitions.

Here's a script to fsck selected partitions every 'N' shutdowns. No more boot delays for fsck (unless something is really wrong

Cheers,

Update1: On my system '/usr/libexec/gam_server' (gamin component used by xfce) prevented /home from being unmounted. I changed

Code:

DISK=/dev/mapper/home; MAX_MCOUNT=1; diskchk

to

Code:

DISK=/dev/mapper/home; MAX_MCOUNT=1; pkill -f gam_server; diskchk

and now my cryptsetup/LUKS home directory gets fscked on shutdown.

Update2: Patch to enhance partition unmount logic.

Code:

--- rc.local_shutdown    	2010-09-21 12:12:53.000000000 -0700
+++ rc.local_shutdown.0.4	2010-09-21 12:20:36.000000000 -0700
@@ -3,6 +3,7 @@
 # Manage large, non-system, ext2/3/4 filesystem checks at
 # shutdown rather than boot.  Tested with Slackware 13.1
 
+# ver 0.4.final: /proc/mounts check
 # ver 0.3.final, em dot lazardo at gmail
 
 function message () {
@@ -16,16 +17,21 @@
 	message "$DISK mount count = ${CUR_MCOUNT[2]}/$MAX_MCOUNT"
 
 	if [ ${CUR_MCOUNT[2]} -gt $MAX_MCOUNT ]; then
-		umount -v $DISK
-		if [ $? -ne 0 ]; then
-			message "cant unmount $DISK (fuser: `fuser -c $DISK`)"
-		else
-			message "starting $DISK fsck"
-			( fsck -yfC $DISK; message "finished $DISK fsck" ) &
-
-			PID[$cnt]=$!
-			((cnt++))
+		grep -q $DISK /proc/mounts
+		if [ $? -eq 0 ]; then
+			umount -v $DISK
+			if [ $? -ne 0 ]; then
+				message "cant unmount $DISK (fuser: `fuser -c $DISK`)"
+				sleep 5
+				return
+			fi
 		fi
+
+		message "starting $DISK fsck"
+		( fsck -yfC $DISK; message "finished $DISK fsck" ) &
+
+		PID[$cnt]=$!
+		((cnt++))
 	fi
 }

Code:

#!/bin/bash
# /etc/rc.d/rc.local_shutdown
# Manage large, non-system, ext2/3/4 filesystem checks at
# shutdown rather than boot.  Tested with Slackware 13.1

# ver 0.3.final, em dot lazardo at gmail

function message () {
	logger -st rc.local_shutdown "$1"
}

function diskchk () {
	# 'tune2fs' and 'fsck' assume ext2/3/4 file system

	CUR_MCOUNT=(`tune2fs -l $DISK | grep '^Mount count:'`)
	message "$DISK mount count = ${CUR_MCOUNT[2]}/$MAX_MCOUNT"

	if [ ${CUR_MCOUNT[2]} -gt $MAX_MCOUNT ]; then
		umount -v $DISK
		if [ $? -ne 0 ]; then
			message "cant unmount $DISK (fuser: `fuser -c $DISK`)"
		else
			message "starting $DISK fsck"
			( fsck -yfC $DISK; message "finished $DISK fsck" ) &

			PID[$cnt]=$!
			((cnt++))
		fi
	fi
}

# Exceptions ###############################################

# reboot
RUN_STAT=( `runlevel` )
case ${RUN_STAT[1]} in
	6|1|S|s)
		message "skipping fsck check (runlevel ${RUN_STAT[1]})"
		exit
		;;
esac

# laptop on single battery
if [ -e /proc/acpi/battery ]; then
	BAT_STAT=`grep discharging /proc/acpi/battery/BAT0/state`
	if [ "$BAT_STAT" != "" ]; then
		message "skipping fsck check (battery)"
		exit
	fi
fi

# shutdown -f
if [ -e /fastboot ]; then
	message "skipping fsck check (shutdown -f)"
	exit
fi

# UPS
# One of these may work, neither was tested.
#
#if [[ -e /etc/apcupsd/powerfail || -e /etc/powerkill ]]; then
#	message "skipping fsck check (UPS)"
#	exit
#fi
#
#if [ -e /etc/powerstatus ]; then
#	UPS_STAT=`grep OK /etc/powerstatus`
#	if [ "$UPS_STAT" != "OK" ]; then
#		message "skipping fsck check (UPS)"
#		exit
#	fi
#fi

# main #####################################################
cnt=0

# Different MAX_MCOUNT values force serialization, identical values run in parallel.
# filesystems active at shutdown will not unmount and are not suitable.

message "Edit DISK and MAX_MCOUNT parameters to match your system and delete this line"

#DISK=/dev/md5; MAX_MCOUNT=13; diskchk		# serial
#DISK=/dev/md6; MAX_MCOUNT=14; diskchk		# serial
#DISK=/dev/sdc1; MAX_MCOUNT=15; diskchk		# parallel
#DISK=/dev/sdd1; MAX_MCOUNT=15; diskchk		# parallel

wait ${PID[*]}

Drakeo · 09-24-2010, 10:27 PM

I just stop running ext2 and ext3 Seems ext4 does my 500 gig drives with in seconds not minutes.
but thanks for the scripts

lazardo · 10-02-2010, 11:31 AM

Quote:

Originally Posted by Drakeo

I just stop running ext2 and ext3 Seems ext4 does my 500 gig drives with in seconds not minutes.
but thanks for the scripts

The script was created to help keep large, actively used media file collections on ext3 until ext4 has a bit more soak time (appologies to Ted Ts'o

.

Cheers,

Richard Cranium · 10-02-2010, 08:51 PM

Why not use xfs instead?

lazardo · 10-03-2010, 12:09 AM

I've seen first hand the ugly underbelly of xfs corruption, performance and recovery issues at a 4 year storage startup stint, and that was with ex-SGI filesystem engineering on board. It's simply not suitable in non-commercial environments and/or without low-level engineering support.

I have a read-intensive, 'smaller numbers of larger files', low risk tolerance, low churn application and a lot of ext2/3 experience which tends to bias choices.

Cheers,