Illumos 5351, 5352 - scrub pauses

5351 scrub goes for an extra second each txg
5352 scrub should pause when there is some dirty data

Author: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Alex Reece <alex.reece@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Richard Elling <richard.elling@richardelling.com>
Approved by: Dan McDonald <danmcd@omniti.com>

References:
  https://www.illumos.org/issues/5351
  https://github.com/illumos/illumos-gate/commit/6f6a76a

Ported-by: Chris Dunlop <chris@onthe.net.au>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #3383

diff --git a/module/zfs/dsl_scan.c b/module/zfs/dsl_scan.c
index 31e693149f6808c3691560cd229e9cd9889c852f..6b9b04a0ee19d0b4a2ec39b12d9fb47b652e552a 100644 (file)
--- a/module/zfs/dsl_scan.c
+++ b/module/zfs/dsl_scan.c
@@ -429,11 +429,14 @@ dsl_scan_sync_state(dsl_scan_t *scn, dmu_tx_t *tx)
            &scn->scn_phys, tx));
 }
 
+extern int zfs_vdev_async_write_active_min_dirty_percent;
+
 static boolean_t
 dsl_scan_check_pause(dsl_scan_t *scn, const zbookmark_phys_t *zb)
 {
        uint64_t elapsed_nanosecs;
        int mintime;
+       int dirty_pct;
 
        /* we never skip user/group accounting objects */
        if (zb && (int64_t)zb->zb_object < 0)
@@ -449,12 +452,28 @@ dsl_scan_check_pause(dsl_scan_t *scn, const zbookmark_phys_t *zb)
        if (zb && zb->zb_level != 0)
                return (B_FALSE);
 
+       /*
+        * We pause if:
+        *  - we have scanned for the maximum time: an entire txg
+        *    timeout (default 5 sec)
+        *  or
+        *  - we have scanned for at least the minimum time (default 1 sec
+        *    for scrub, 3 sec for resilver), and either we have sufficient
+        *    dirty data that we are starting to write more quickly
+        *    (default 30%), or someone is explicitly waiting for this txg
+        *    to complete.
+        *  or
+        *  - the spa is shutting down because this pool is being exported
+        *    or the machine is rebooting.
+        */
        mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
            zfs_resilver_min_time_ms : zfs_scan_min_time_ms;
        elapsed_nanosecs = gethrtime() - scn->scn_sync_start_time;
-       if (elapsed_nanosecs / NANOSEC > zfs_txg_timeout ||
+       dirty_pct = scn->scn_dp->dp_dirty_total * 100 / zfs_dirty_data_max;
+       if (elapsed_nanosecs / NANOSEC >= zfs_txg_timeout ||
            (NSEC2MSEC(elapsed_nanosecs) > mintime &&
-           txg_sync_waiting(scn->scn_dp)) ||
+           (txg_sync_waiting(scn->scn_dp) ||
+           dirty_pct >= zfs_vdev_async_write_active_min_dirty_percent)) ||
            spa_shutting_down(scn->scn_dp->dp_spa)) {
                if (zb) {
                        dprintf("pausing at bookmark %llx/%llx/%llx/%llx\n",