md-rebuild-policy.patch

lustre 1.6.5 source code

diff -pur linux-2.6.18-53.orig/drivers/md/md.c linux-2.6.18-53/drivers/md/md.c
--- linux-2.6.18-53.orig/drivers/md/md.c	2008-02-13 17:34:25.000000000 +0800
+++ linux-2.6.18-53/drivers/md/md.c	2008-02-13 17:39:28.000000000 +0800
@@ -90,6 +90,8 @@ static void md_print_devices(void);
 
 static int sysctl_speed_limit_min = 1000;
 static int sysctl_speed_limit_max = 200000;
+static int sysctl_rebuild_window_size = 256;
+static int sysctl_disk_idle_size = 4096;
 static inline int speed_min(mddev_t *mddev)
 {
 	return mddev->sync_speed_min ?
@@ -121,6 +123,22 @@ static ctl_table raid_table[] = {
 		.mode		= S_IRUGO|S_IWUSR,
 		.proc_handler	= &proc_dointvec,
 	},
+	{
+		.ctl_name	= DEV_RAID_REBUILD_WINDOW,
+		.procname	= "rebuild_window_size",
+		.data		= &sysctl_rebuild_window_size,
+		.maxlen		= sizeof(int),
+		.mode		= S_IRUGO|S_IWUSR,
+		.proc_handler	= &proc_dointvec,
+	},
+	{
+		.ctl_name	= DEV_RAID_DISK_IDLE_SIZE,
+		.procname	= "disk_idle_size",
+		.data		= &sysctl_disk_idle_size,
+		.maxlen		= sizeof(int),
+		.mode		= S_IRUGO|S_IWUSR,
+		.proc_handler	= &proc_dointvec,
+	},
 	{ .ctl_name = 0 }
 };
 
@@ -4980,14 +4998,15 @@ static int is_mddev_idle(mddev_t *mddev)
 	mdk_rdev_t * rdev;
 	struct list_head *tmp;
 	int idle;
-	unsigned long curr_events;
+	unsigned long rw, sync;
 
 	idle = 1;
 	ITERATE_RDEV(mddev,rdev,tmp) {
 		struct gendisk *disk = rdev->bdev->bd_contains->bd_disk;
-		curr_events = disk_stat_read(disk, sectors[0]) + 
-				disk_stat_read(disk, sectors[1]) - 
-				atomic_read(&disk->sync_io);
+
+		rw = disk_stat_read(disk, sectors[READ])+disk_stat_read(disk, sectors[WRITE]);
+		sync = atomic_read(&disk->sync_io);
+
 		/* The difference between curr_events and last_events
 		 * will be affected by any new non-sync IO (making
 		 * curr_events bigger) and any difference in the amount of
@@ -5001,9 +5020,9 @@ static int is_mddev_idle(mddev_t *mddev)
 		 *
 		 * Note: the following is an unsigned comparison.
 		 */
-		if ((curr_events - rdev->last_events + 4096) > 8192) {
-			rdev->last_events = curr_events;
+		if (rw - rdev->last_events > sync + sysctl_disk_idle_size) {
 			idle = 0;
+			rdev->last_events = rw - sync;
 		}
 	}
 	return idle;
@@ -5069,8 +5088,7 @@ static DECLARE_WAIT_QUEUE_HEAD(resync_wa
 void md_do_sync(mddev_t *mddev)
 {
 	mddev_t *mddev2;
-	unsigned int currspeed = 0,
-		 window;
+	unsigned int currspeed = 0;
 	sector_t max_sectors,j, io_sectors;
 	unsigned long mark[SYNC_MARKS];
 	sector_t mark_cnt[SYNC_MARKS];
@@ -5190,9 +5208,8 @@ void md_do_sync(mddev_t *mddev)
 	/*
 	 * Tune reconstruction:
 	 */
-	window = 32*(PAGE_SIZE/512);
 	printk(KERN_INFO "md: using %dk window, over a total of %llu blocks.\n",
-		window/2,(unsigned long long) max_sectors/2);
+		sysctl_rebuild_window_size/2,(unsigned long long) max_sectors/2);
 
 	atomic_set(&mddev->recovery_active, 0);
 	init_waitqueue_head(&mddev->recovery_wait);
@@ -5230,7 +5247,7 @@ void md_do_sync(mddev_t *mddev)
 			 */
 			md_new_event(mddev);
 
-		if (last_check + window > io_sectors || j == max_sectors)
+		if (last_check + sysctl_rebuild_window_size > io_sectors || j == max_sectors)
 			continue;
 
 		last_check = io_sectors;
@@ -5251,7 +5268,6 @@ void md_do_sync(mddev_t *mddev)
 			last_mark = next;
 		}
 
-
 		if (kthread_should_stop()) {
 			/*
 			 * got a signal, exit.
@@ -5275,10 +5291,16 @@ void md_do_sync(mddev_t *mddev)
 
 		currspeed = ((unsigned long)(io_sectors-mddev->resync_mark_cnt))/2
 			/((jiffies-mddev->resync_mark)/HZ +1) +1;
-
 		if (currspeed > speed_min(mddev)) {
 			if ((currspeed > speed_max(mddev)) ||
 					!is_mddev_idle(mddev)) {
+				static unsigned long next_report;
+				if (time_after(jiffies, next_report)) {
+					printk(KERN_INFO "md: rebuild %s throttled due to IO\n",
+						mdname(mddev));
+					/* once per 10 minutes */
+					next_report = jiffies + 600 * HZ;
+				}
 				msleep(500);
 				goto repeat;
 			}
diff -pur linux-2.6.18-53.orig/include/linux/sysctl.h linux-2.6.18-53/include/linux/sysctl.h
--- linux-2.6.18-53.orig/include/linux/sysctl.h	2008-02-13 17:35:25.000000000 +0800
+++ linux-2.6.18-53/include/linux/sysctl.h	2008-02-13 17:36:22.000000000 +0800
@@ -903,7 +903,9 @@ enum {
 /* /proc/sys/dev/raid */
 enum {
 	DEV_RAID_SPEED_LIMIT_MIN=1,
-	DEV_RAID_SPEED_LIMIT_MAX=2
+	DEV_RAID_SPEED_LIMIT_MAX=2,
+	DEV_RAID_REBUILD_WINDOW=3,
+	DEV_RAID_DISK_IDLE_SIZE=4
 };
 
 /* /proc/sys/dev/parport/default */