OpenZFS 9102 - zfs should be able to initialize storage devices

PROBLEM
========

The first access to a block incurs a performance penalty on some platforms
(e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are
"thick provisioned", where supported by the platform (VMware). This can
create a large delay in getting a new virtual machines up and running (or
adding storage to an existing Engine). If the thick provision step is
omitted, write performance will be suboptimal until all blocks on the LUN
have been written.

SOLUTION
=========

This feature introduces a way to 'initialize' the disks at install or in the
background to make sure we don't incur this first read penalty.

When an entire LUN is added to ZFS, we make all space available immediately,
and allow ZFS to find unallocated space and zero it out. This works with
concurrent writes to arbitrary offsets, ensuring that we don't zero out
something that has been (or is in the middle of being) written. This scheme
can also be applied to existing pools (affecting only free regions on the
vdev). Detailed design:
        - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...]
                - start, suspend, or cancel initialization
        - Creates new open-context thread for each vdev
        - Thread iterates through all metaslabs in this vdev
        - Each metaslab:
                - select a metaslab
                - load the metaslab
                - mark the metaslab as being zeroed
                - walk all free ranges within that metaslab and translate
                  them to ranges on the leaf vdev
                - issue a "zeroing" I/O on the leaf vdev that corresponds to
                  a free range on the metaslab we're working on
                - continue until all free ranges for this metaslab have been
                  "zeroed"
                - reset/unmark the metaslab being zeroed
                - if more metaslabs exist, then repeat above tasks.
                - if no more metaslabs, then we're done.

        - progress for the initialization is stored on-disk in the vdev’s
          leaf zap object. The following information is stored:
                - the last offset that has been initialized
                - the state of the initialization process (i.e. active,
                  suspended, or canceled)
                - the start time for the initialization

        - progress is reported via the zpool status command and shows
          information for each of the vdevs that are initializing

Porting notes:
- Added zfs_initialize_value module parameter to set the pattern
  written by "zpool initialize".
- Added zfs_vdev_{initializing,removal}_{min,max}_active module options.

Authored by: George Wilson <[email protected]>
Reviewed by: John Wren Kennedy <[email protected]>
Reviewed by: Matthew Ahrens <[email protected]>
Reviewed by: Pavel Zakharov <[email protected]>
Reviewed by: Prakash Surya <[email protected]>
Reviewed by: loli10K <[email protected]>
Reviewed by: Brian Behlendorf <[email protected]>
Approved by: Richard Lowe <[email protected]>
Signed-off-by: Tim Chase <[email protected]>
Ported-by: Tim Chase <[email protected]>

OpenZFS-issue: https://www.illumos.org/issues/9102
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb
Closes #8230

1 files changed, 44 insertions, 2 deletions

diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
index 8273e7907..f808f8ee7 100644
--- a/module/zfs/vdev.c
+++ b/module/zfs/vdev.c
@@ -50,6 +50,7 @@
 #include <sys/zil.h>
 #include <sys/dsl_scan.h>
 #include <sys/abd.h>
+#include <sys/vdev_initialize.h>
 #include <sys/zvol.h>
 #include <sys/zfs_ratelimit.h>
 
@@ -212,6 +213,14 @@ vdev_getops(const char *type)
 	return (ops);
 }
 
+/* ARGSUSED */
+void
+vdev_default_xlate(vdev_t *vd, const range_seg_t *in, range_seg_t *res)
+{
+	res->rs_start = in->rs_start;
+	res->rs_end = in->rs_end;
+}
+
 /*
  * Derive the enumerated alloction bias from string input.
  * String origin is either the per-vdev zap or zpool(1M).
@@ -526,6 +535,10 @@ vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops)
 	mutex_init(&vd->vdev_probe_lock, NULL, MUTEX_DEFAULT, NULL);
 	mutex_init(&vd->vdev_queue_lock, NULL, MUTEX_DEFAULT, NULL);
 	mutex_init(&vd->vdev_scan_io_queue_lock, NULL, MUTEX_DEFAULT, NULL);
+	mutex_init(&vd->vdev_initialize_lock, NULL, MUTEX_DEFAULT, NULL);
+	mutex_init(&vd->vdev_initialize_io_lock, NULL, MUTEX_DEFAULT, NULL);
+	cv_init(&vd->vdev_initialize_cv, NULL, CV_DEFAULT, NULL);
+	cv_init(&vd->vdev_initialize_io_cv, NULL, CV_DEFAULT, NULL);
 
 	for (int t = 0; t < DTL_TYPES; t++) {
 		vd->vdev_dtl[t] = range_tree_create(NULL, NULL);
@@ -850,6 +863,7 @@ void
 vdev_free(vdev_t *vd)
 {
 	spa_t *spa = vd->vdev_spa;
+	ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
 
 	/*
 	 * Scan queues are normally destroyed at the end of a scan. If the
@@ -880,6 +894,7 @@ vdev_free(vdev_t *vd)
 
 	ASSERT(vd->vdev_child == NULL);
 	ASSERT(vd->vdev_guid_sum == vd->vdev_guid);
+	ASSERT(vd->vdev_initialize_thread == NULL);
 
 	/*
 	 * Discard allocation state.
@@ -957,6 +972,10 @@ vdev_free(vdev_t *vd)
 	mutex_destroy(&vd->vdev_stat_lock);
 	mutex_destroy(&vd->vdev_probe_lock);
 	mutex_destroy(&vd->vdev_scan_io_queue_lock);
+	mutex_destroy(&vd->vdev_initialize_lock);
+	mutex_destroy(&vd->vdev_initialize_io_lock);
+	cv_destroy(&vd->vdev_initialize_io_cv);
+	cv_destroy(&vd->vdev_initialize_cv);
 
 	zfs_ratelimit_fini(&vd->vdev_delay_rl);
 	zfs_ratelimit_fini(&vd->vdev_checksum_rl);
@@ -3207,7 +3226,8 @@ vdev_sync_done(vdev_t *vd, uint64_t txg)
 
 	ASSERT(vdev_is_concrete(vd));
 
-	while ((msp = txg_list_remove(&vd->vdev_ms_list, TXG_CLEAN(txg))))
+	while ((msp = txg_list_remove(&vd->vdev_ms_list, TXG_CLEAN(txg)))
+	    != NULL)
 		metaslab_sync_done(msp, txg);
 
 	if (reassess)
@@ -3458,6 +3478,15 @@ vdev_online(spa_t *spa, uint64_t guid, uint64_t flags, vdev_state_t *newstate)
 		spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
 	}
 
+	/* Restart initializing if necessary */
+	mutex_enter(&vd->vdev_initialize_lock);
+	if (vdev_writeable(vd) &&
+	    vd->vdev_initialize_thread == NULL &&
+	    vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE) {
+		(void) vdev_initialize(vd);
+	}
+	mutex_exit(&vd->vdev_initialize_lock);
+
 	if (wasoffline ||
 	    (oldstate < VDEV_STATE_DEGRADED &&
 	    vd->vdev_state >= VDEV_STATE_DEGRADED))
@@ -3848,9 +3877,22 @@ vdev_get_stats_ex(vdev_t *vd, vdev_stat_t *vs, vdev_stat_ex_t *vsx)
 		vs->vs_timestamp = gethrtime() - vs->vs_timestamp;
 		vs->vs_state = vd->vdev_state;
 		vs->vs_rsize = vdev_get_min_asize(vd);
-		if (vd->vdev_ops->vdev_op_leaf)
+		if (vd->vdev_ops->vdev_op_leaf) {
 			vs->vs_rsize += VDEV_LABEL_START_SIZE +
 			    VDEV_LABEL_END_SIZE;
+			/*
+			 * Report intializing progress. Since we don't
+			 * have the initializing locks held, this is only
+			 * an estimate (although a fairly accurate one).
+			 */
+			vs->vs_initialize_bytes_done =
+			    vd->vdev_initialize_bytes_done;
+			vs->vs_initialize_bytes_est =
+			    vd->vdev_initialize_bytes_est;
+			vs->vs_initialize_state = vd->vdev_initialize_state;
+			vs->vs_initialize_action_time =
+			    vd->vdev_initialize_action_time;
+		}
 		/*
 		 * Report expandable space on top-level, non-auxillary devices
 		 * only. The expandable space is reported in terms of metaslab