Change checksum & IO delay ratelimit values

Change checksum & IO delay ratelimit thresholds from 5/sec to 20/sec.
This allows zed to actually trigger if a bunch of these events arrive in
a short period of time (zed has a threshold of 10 events in 10 sec).
Previously, if you had, say, 100 checksum errors in 1 sec, it would get
ratelimited to 5/sec which wouldn't trigger zed to fault the drive.

Also, convert the checksum and IO delay thresholds to module params for
easy testing.

Reviewed-by: loli10K <ezomori.nozomu@gmail.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov>
Signed-off-by: Tony Hutter <hutter2@llnl.gov>
Closes #7252

include/sys/vdev_impl.h

@@ -262,8 +262,6 @@ struct vdev {
 	 * We rate limit ZIO delay and ZIO checksum events, since they
 	 * can flood ZED with tons of events when a drive is acting up.
 	 */
-#define	DELAYS_PER_SECOND 5
-#define	CHECKSUMS_PER_SECOND 5
 	zfs_ratelimit_t vdev_delay_rl;
 	zfs_ratelimit_t vdev_checksum_rl;
 };

include/sys/zfs_ratelimit.h

@@ -25,13 +25,19 @@
 typedef struct {
 	hrtime_t start;
 	unsigned int count;
-	unsigned int burst;		/* Number to allow per interval */
+
-	unsigned int interval;		/* Interval length in seconds */
+	/*
+	 * Pointer to number of events per interval.  We do this to
+	 * allow the burst to be a (changeable) module parameter.
+	 */
+	unsigned int *burst;
+
+	unsigned int interval;	/* Interval length in seconds */
 	kmutex_t lock;
 } zfs_ratelimit_t;
 
 int zfs_ratelimit(zfs_ratelimit_t *rl);
-void zfs_ratelimit_init(zfs_ratelimit_t *rl, unsigned int burst,
+void zfs_ratelimit_init(zfs_ratelimit_t *rl, unsigned int *burst,
     unsigned int interval);
 void zfs_ratelimit_fini(zfs_ratelimit_t *rl);
 

man/man5/zfs-module-parameters.5

@@ -753,6 +753,19 @@ Disable pool import at module load by ignoring the cache file (typically \fB/etc
 Use \fB1\fR for yes (default) and \fB0\fR for no.
 .RE
 
+.sp
+.ne 2
+.na
+\fBzfs_checksums_per_second\fR (int)
+.ad
+.RS 12n
+Rate limit checksum events to this many per second.  Note that this should
+not be set below the zed thresholds (currently 10 checksums over 10 sec)
+or else zed may not trigger any action.
+.sp
+Default value: 20
+.RE
+
 .sp
 .ne 2
 .na
@@ -929,6 +942,17 @@ Note: \fBzfs_delay_scale\fR * \fBzfs_dirty_data_max\fR must be < 2^64.
 Default value: \fB500,000\fR.
 .RE
 
+.sp
+.ne 2
+.na
+\fBzfs_delays_per_second\fR (int)
+.ad
+.RS 12n
+Rate limit IO delay events to this many per second.
+.sp
+Default value: 20
+.RE
+
 .sp
 .ne 2
 .na

module/zfs/vdev.c

@@ -56,6 +56,16 @@
  */
 int metaslabs_per_vdev = 200;
 
+/*
+ * Rate limit delay events to this many IO delays per second.
+ */
+unsigned int zfs_delays_per_second = 20;
+
+/*
+ * Rate limit checksum events after this many checksum errors per second.
+ */
+unsigned int zfs_checksums_per_second = 20;
+
 /*
  * Virtual device management.
  */
@@ -351,8 +361,8 @@ vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops)
 	 * and checksum events so that we don't overwhelm ZED with thousands
 	 * of events when a disk is acting up.
 	 */
-	zfs_ratelimit_init(&vd->vdev_delay_rl, DELAYS_PER_SECOND, 1);
+	zfs_ratelimit_init(&vd->vdev_delay_rl, &zfs_delays_per_second, 1);
-	zfs_ratelimit_init(&vd->vdev_checksum_rl, CHECKSUMS_PER_SECOND, 1);
+	zfs_ratelimit_init(&vd->vdev_checksum_rl, &zfs_checksums_per_second, 1);
 
 	list_link_init(&vd->vdev_config_dirty_node);
 	list_link_init(&vd->vdev_state_dirty_node);
@@ -3752,5 +3762,14 @@ module_param(metaslabs_per_vdev, int, 0644);
 MODULE_PARM_DESC(metaslabs_per_vdev,
 	"Divide added vdev into approximately (but no more than) this number "
 	"of metaslabs");
+
+module_param(zfs_delays_per_second, uint, 0644);
+MODULE_PARM_DESC(zfs_delays_per_second, "Rate limit delay events to this many "
+	"IO delays per second");
+
+module_param(zfs_checksums_per_second, uint, 0644);
+	MODULE_PARM_DESC(zfs_checksums_per_second, "Rate limit checksum events "
+	"to this many checksum errors per second (do not set below zed"
+	"threshold).");
 /* END CSTYLED */
 #endif

module/zfs/zfs_ratelimit.c

@@ -33,7 +33,7 @@
  * interval:	Interval time in seconds
  */
 void
-zfs_ratelimit_init(zfs_ratelimit_t *rl, unsigned int burst,
+zfs_ratelimit_init(zfs_ratelimit_t *rl, unsigned int *burst,
     unsigned int interval)
 {
 	rl->count = 0;
@@ -89,7 +89,7 @@ zfs_ratelimit(zfs_ratelimit_t *rl)
 		rl->start = now;
 		rl->count = 0;
 	} else {
-		if (rl->count >= rl->burst) {
+		if (rl->count >= *rl->burst) {
 			error = 0; /* We're ratelimiting */
 		}
 	}

tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh

@@ -68,14 +68,8 @@ for type in "mirror" "raidz" "raidz2"; do
 	log_must dd if=/dev/urandom of=$TESTFILE bs=1M count=16
 
 	# 4. Inject CHECKSUM ERRORS on read with a zinject error handler
-	# NOTE: checksum events are ratelimited to max 5 per second, ZED needs
-	#       10 to kick in a spare
 	log_must zinject -d $FAULT_FILE -e corrupt -f 50 -T read $TESTPOOL
 	log_must cp $TESTFILE /dev/null
-	log_must sleep 1
-	log_must cp $TESTFILE /dev/null
-	log_must sleep 1
-	log_must cp $TESTFILE /dev/null
 
 	# 5. Verify the ZED kicks in a hot spare and expected pool/device status
 	log_note "Wait for ZED to auto-spare"