Improve resilver ETAs

When resilvering the estimated time remaining is calculated using
the average issue rate over the current pass.  Where the current
pass starts when a scan was started, or restarted, if the pool
was exported/imported.

For dRAID pools in particular this can result in wildly optimistic
estimates since the issue rate will be very high while scanning
when non-degraded regions of the pool are scanned.  Once repair
I/O starts being issued performance drops to a realistic number
but the estimated performance is still significantly skewed.

To address this we redefine a pass such that it starts after a
scanning phase completes so the issue rate is more reflective of
recent performance.  Additionally, the zfs_scan_report_txgs
module option can be set to reset the pass statistics more often.

Reviewed-by: Akash B <[email protected]>
Reviewed-by: Tony Hutter <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Closes openzfs#14410

cmd/zpool/zpool_main.c

@@ -7524,30 +7524,31 @@ print_scan_scrub_resilver_status(pool_scan_stat_t *ps)
 
 	zfs_nicebytes(ps->pss_processed, processed_buf, sizeof (processed_buf));
 
-	assert(ps->pss_func == POOL_SCAN_SCRUB ||
-	    ps->pss_func == POOL_SCAN_RESILVER);
+	int is_resilver = ps->pss_func == POOL_SCAN_RESILVER;
+	int is_scrub = ps->pss_func == POOL_SCAN_SCRUB;
+	assert(is_resilver || is_scrub);
 
 	/* Scan is finished or canceled. */
 	if (ps->pss_state == DSS_FINISHED) {
 		secs_to_dhms(end - start, time_buf);
 
-		if (ps->pss_func == POOL_SCAN_SCRUB) {
+		if (is_scrub) {
 			(void) printf(gettext("scrub repaired %s "
 			    "in %s with %llu errors on %s"), processed_buf,
 			    time_buf, (u_longlong_t)ps->pss_errors,
 			    ctime(&end));
-		} else if (ps->pss_func == POOL_SCAN_RESILVER) {
+		} else if (is_resilver) {
 			(void) printf(gettext("resilvered %s "
 			    "in %s with %llu errors on %s"), processed_buf,
 			    time_buf, (u_longlong_t)ps->pss_errors,
 			    ctime(&end));
 		}
 		return;
 	} else if (ps->pss_state == DSS_CANCELED) {
-		if (ps->pss_func == POOL_SCAN_SCRUB) {
+		if (is_scrub) {
 			(void) printf(gettext("scrub canceled on %s"),
 			    ctime(&end));
-		} else if (ps->pss_func == POOL_SCAN_RESILVER) {
+		} else if (is_resilver) {
 			(void) printf(gettext("resilver canceled on %s"),
 			    ctime(&end));
 		}
@@ -7557,7 +7558,7 @@ print_scan_scrub_resilver_status(pool_scan_stat_t *ps)
 	assert(ps->pss_state == DSS_SCANNING);
 
 	/* Scan is in progress. Resilvers can't be paused. */
-	if (ps->pss_func == POOL_SCAN_SCRUB) {
+	if (is_scrub) {
 		if (pause == 0) {
 			(void) printf(gettext("scrub in progress since %s"),
 			    ctime(&start));
@@ -7567,7 +7568,7 @@ print_scan_scrub_resilver_status(pool_scan_stat_t *ps)
 			(void) printf(gettext("\tscrub started on %s"),
 			    ctime(&start));
 		}
-	} else if (ps->pss_func == POOL_SCAN_RESILVER) {
+	} else if (is_resilver) {
 		(void) printf(gettext("resilver in progress since %s"),
 		    ctime(&start));
 	}
@@ -7609,17 +7610,27 @@ print_scan_scrub_resilver_status(pool_scan_stat_t *ps)
 		    scanned_buf, issued_buf, total_buf);
 	}
 
-	if (ps->pss_func == POOL_SCAN_RESILVER) {
+	if (is_resilver) {
 		(void) printf(gettext("\t%s resilvered, %.2f%% done"),
 		    processed_buf, 100 * fraction_done);
-	} else if (ps->pss_func == POOL_SCAN_SCRUB) {
+	} else if (is_scrub) {
 		(void) printf(gettext("\t%s repaired, %.2f%% done"),
 		    processed_buf, 100 * fraction_done);
 	}
 
 	if (pause == 0) {
+		/*
+		 * Only provide an estimate iff:
+		 * 1) the time remaining is valid, and
+		 * 2) the issue rate exceeds 10 MB/s, and
+		 * 3) it's either:
+		 *    a) a resilver which has started repairs, or
+		 *    b) a scrub which has entered the issue phase.
+		 */
 		if (total_secs_left != UINT64_MAX &&
-		    issue_rate >= 10 * 1024 * 1024) {
+		    issue_rate >= 10 * 1024 * 1024 &&
+		    ((is_resilver && ps->pss_processed > 0) ||
+		    (is_scrub && issued > 0))) {
 			(void) printf(gettext(", %s to go\n"), time_buf);
 		} else {
 			(void) printf(gettext(", no estimated "

man/man4/zfs.4

@@ -1890,6 +1890,13 @@ I/O.
 In this case (unless the metadata scan is done) we stop issuing verification I/O
 and start scanning metadata again until we get to the hard limit.
 .
+.It Sy zfs_scan_report_txgs Ns = Ns Sy 0 Ns | Ns 1 Pq uint
+When reporting resilver throughput and estimated completion time use the
+performance observed over roughly the last
+.Sy zfs_scan_report_txgs
+TXGs.
+When set to zero performance is calculated over the time between checkpoints.
+.
 .It Sy zfs_scan_strict_mem_lim Ns = Ns Sy 0 Ns | Ns 1 Pq int
 Enforce tight memory limits on pool scans when a sequential scan is in progress.
 When disabled, the memory limit may be exceeded by fast disks.

module/zfs/dsl_scan.c

@@ -131,6 +131,15 @@ static uint64_t dsl_scan_count_data_disks(vdev_t *vd);
 extern uint_t zfs_vdev_async_write_active_min_dirty_percent;
 static int zfs_scan_blkstats = 0;
 
+/*
+ * 'zpool status' uses bytes processed per pass to report throughput and
+ * estimate time remaining.  We define a pass to start when the scanning
+ * phase completes for a sequential resilver.  Optionally, this value
+ * may be used to reset the pass statistics every N txgs to provide an
+ * estimated completion time based on currently observed performance.
+ */
+static uint_t zfs_scan_report_txgs = 0;
+
 /*
  * By default zfs will check to ensure it is not over the hard memory
  * limit before each txg. If finer-grained control of this is needed
@@ -604,6 +613,8 @@ dsl_scan_init(dsl_pool_t *dp, uint64_t txg)
 	}
 
 	spa_scan_stat_init(spa);
+	vdev_scan_stat_init(spa->spa_root_vdev);
+
 	return (0);
 }
 
@@ -763,6 +774,7 @@ dsl_scan_setup_sync(void *arg, dmu_tx_t *tx)
 	scn->scn_last_checkpoint = 0;
 	scn->scn_checkpointing = B_FALSE;
 	spa_scan_stat_init(spa);
+	vdev_scan_stat_init(spa->spa_root_vdev);
 
 	if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
 		scn->scn_phys.scn_ddt_class_max = zfs_scrub_ddt_class_max;
@@ -3686,6 +3698,16 @@ dsl_scan_sync(dsl_pool_t *dp, dmu_tx_t *tx)
 		return;
 	}
 
+	/*
+	 * Disabled by default, set zfs_scan_report_txgs to report
+	 * average performance over the last zfs_scan_report_txgs TXGs.
+	 */
+	if (!dsl_scan_is_paused_scrub(scn) && zfs_scan_report_txgs != 0 &&
+	    tx->tx_txg % zfs_scan_report_txgs == 0) {
+		scn->scn_issued_before_pass += spa->spa_scan_pass_issued;
+		spa_scan_stat_init(spa);
+	}
+
 	/*
 	 * It is possible to switch from unsorted to sorted at any time,
 	 * but afterwards the scan will remain sorted unless reloaded from
@@ -3814,6 +3836,9 @@ dsl_scan_sync(dsl_pool_t *dp, dmu_tx_t *tx)
 			if (scn->scn_is_sorted) {
 				scn->scn_checkpointing = B_TRUE;
 				scn->scn_clearing = B_TRUE;
+				scn->scn_issued_before_pass +=
+				    spa->spa_scan_pass_issued;
+				spa_scan_stat_init(spa);
 			}
 			zfs_dbgmsg("scan complete for %s txg %llu",
 			    spa->spa_name,
@@ -4541,5 +4566,8 @@ ZFS_MODULE_PARAM(zfs, zfs_, scan_strict_mem_lim, INT, ZMOD_RW,
 ZFS_MODULE_PARAM(zfs, zfs_, scan_fill_weight, UINT, ZMOD_RW,
 	"Tunable to adjust bias towards more filled segments during scans");
 
+ZFS_MODULE_PARAM(zfs, zfs_, scan_report_txgs, UINT, ZMOD_RW,
+	"Tunable to report resilver performance over the last N txgs");
+
 ZFS_MODULE_PARAM(zfs, zfs_, resilver_disable_defer, INT, ZMOD_RW,
 	"Process all resilvers immediately");

module/zfs/spa_misc.c

@@ -2556,7 +2556,6 @@ spa_scan_stat_init(spa_t *spa)
 	spa->spa_scan_pass_scrub_spent_paused = 0;
 	spa->spa_scan_pass_exam = 0;
 	spa->spa_scan_pass_issued = 0;
-	vdev_scan_stat_init(spa->spa_root_vdev);
 }
 
 /*