Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Set aside a metaslab for ZIL blocks #11389

Merged
merged 1 commit into from
Jan 21, 2021
Merged

Set aside a metaslab for ZIL blocks #11389

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 19 additions & 2 deletions cmd/zdb/zdb.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5844,6 +5844,7 @@ zdb_leak_init_prepare_indirect_vdevs(spa_t *spa, zdb_cb_t *zcb)
* metaslabs. We want to set them up for
* zio_claim().
*/
vdev_metaslab_group_create(vd);
VERIFY0(vdev_metaslab_init(vd, 0));

vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
Expand Down Expand Up @@ -5882,6 +5883,7 @@ zdb_leak_init(spa_t *spa, zdb_cb_t *zcb)
*/
spa->spa_normal_class->mc_ops = &zdb_metaslab_ops;
spa->spa_log_class->mc_ops = &zdb_metaslab_ops;
spa->spa_embedded_log_class->mc_ops = &zdb_metaslab_ops;

zcb->zcb_vd_obsolete_counts =
umem_zalloc(rvd->vdev_children * sizeof (uint32_t *),
Expand Down Expand Up @@ -6015,15 +6017,16 @@ zdb_leak_fini(spa_t *spa, zdb_cb_t *zcb)
vdev_t *rvd = spa->spa_root_vdev;
for (unsigned c = 0; c < rvd->vdev_children; c++) {
vdev_t *vd = rvd->vdev_child[c];
metaslab_group_t *mg __maybe_unused = vd->vdev_mg;

if (zcb->zcb_vd_obsolete_counts[c] != NULL) {
leaks |= zdb_check_for_obsolete_leaks(vd, zcb);
}

for (uint64_t m = 0; m < vd->vdev_ms_count; m++) {
metaslab_t *msp = vd->vdev_ms[m];
ASSERT3P(mg, ==, msp->ms_group);
ASSERT3P(msp->ms_group, ==, (msp->ms_group->mg_class ==
spa_embedded_log_class(spa)) ?
vd->vdev_log_mg : vd->vdev_mg);

/*
* ms_allocatable has been overloaded
Expand Down Expand Up @@ -6230,6 +6233,8 @@ dump_block_stats(spa_t *spa)
zcb.zcb_totalasize = metaslab_class_get_alloc(spa_normal_class(spa));
zcb.zcb_totalasize += metaslab_class_get_alloc(spa_special_class(spa));
zcb.zcb_totalasize += metaslab_class_get_alloc(spa_dedup_class(spa));
zcb.zcb_totalasize +=
metaslab_class_get_alloc(spa_embedded_log_class(spa));
zcb.zcb_start = zcb.zcb_lastprint = gethrtime();
err = traverse_pool(spa, 0, flags, zdb_blkptr_cb, &zcb);

Expand Down Expand Up @@ -6277,6 +6282,7 @@ dump_block_stats(spa_t *spa)

total_alloc = norm_alloc +
metaslab_class_get_alloc(spa_log_class(spa)) +
metaslab_class_get_alloc(spa_embedded_log_class(spa)) +
metaslab_class_get_alloc(spa_special_class(spa)) +
metaslab_class_get_alloc(spa_dedup_class(spa)) +
get_unflushed_alloc_space(spa);
Expand Down Expand Up @@ -6344,6 +6350,17 @@ dump_block_stats(spa_t *spa)
100.0 * alloc / space);
}

if (spa_embedded_log_class(spa)->mc_allocator[0].mca_rotor != NULL) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is the selection of the embedded_log_class dynamic? (i.e. at each spa_import?). I'm trying to understand how zdb instance of the pool knows which ms to use for embedded log and if it can be different from what the runtime pool is using?

If zdb can know the actual ms, then a follow on zdb change could augment the zdb -m output by tagging the embedded log ms.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, the metaslab is selected when opening the pool. So zdb wouldn't necessarily select the same metaslab as was used recently (or is currently in use, if the pool is imported while running zdb). But typically, on pools that have moderate or higher fullness or fragmentation, the zil metaslab will really stick out as it will be the only one that's (nearly) empty and unfragmented.

uint64_t alloc = metaslab_class_get_alloc(
spa_embedded_log_class(spa));
uint64_t space = metaslab_class_get_space(
spa_embedded_log_class(spa));

(void) printf("\t%-16s %14llu used: %5.2f%%\n",
"Embedded log class", (u_longlong_t)alloc,
100.0 * alloc / space);
}

for (i = 0; i < NUM_BP_EMBEDDED_TYPES; i++) {
if (zcb.zcb_embedded_blocks[i] == 0)
continue;
Expand Down
3 changes: 0 additions & 3 deletions include/sys/dmu.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -142,9 +142,6 @@ typedef enum dmu_object_byteswap {
#define DMU_OT_IS_DDT(ot) \
((ot) == DMU_OT_DDT_ZAP)

#define DMU_OT_IS_ZIL(ot) \
((ot) == DMU_OT_INTENT_LOG)

/* Note: ztest uses DMU_OT_UINT64_OTHER as a proxy for file blocks */
#define DMU_OT_IS_FILE(ot) \
((ot) == DMU_OT_PLAIN_FILE_CONTENTS || (ot) == DMU_OT_UINT64_OTHER)
Expand Down
1 change: 1 addition & 0 deletions include/sys/spa.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1047,6 +1047,7 @@ extern uint64_t spa_version(spa_t *spa);
extern boolean_t spa_deflate(spa_t *spa);
extern metaslab_class_t *spa_normal_class(spa_t *spa);
extern metaslab_class_t *spa_log_class(spa_t *spa);
extern metaslab_class_t *spa_embedded_log_class(spa_t *spa);
extern metaslab_class_t *spa_special_class(spa_t *spa);
extern metaslab_class_t *spa_dedup_class(spa_t *spa);
extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size,
Expand Down
1 change: 1 addition & 0 deletions include/sys/spa_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -226,6 +226,7 @@ struct spa {
boolean_t spa_is_exporting; /* true while exporting pool */
metaslab_class_t *spa_normal_class; /* normal data class */
metaslab_class_t *spa_log_class; /* intent log data class */
metaslab_class_t *spa_embedded_log_class; /* log on normal vdevs */
metaslab_class_t *spa_special_class; /* special allocation class */
metaslab_class_t *spa_dedup_class; /* dedup allocation class */
uint64_t spa_first_txg; /* first txg after spa_open() */
Expand Down
4 changes: 4 additions & 0 deletions include/sys/vdev.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,6 +33,7 @@
#include <sys/zio.h>
#include <sys/dmu.h>
#include <sys/space_map.h>
#include <sys/metaslab.h>
#include <sys/fs/zfs.h>

#ifdef __cplusplus
Expand Down Expand Up @@ -113,6 +114,9 @@ extern void vdev_xlate_walk(vdev_t *vd, const range_seg64_t *logical_rs,
vdev_xlate_func_t *func, void *arg);

extern void vdev_get_stats_ex(vdev_t *vd, vdev_stat_t *vs, vdev_stat_ex_t *vsx);

extern metaslab_group_t *vdev_get_mg(vdev_t *vd, metaslab_class_t *mc);

extern void vdev_get_stats(vdev_t *vd, vdev_stat_t *vs);
extern void vdev_clear_stats(vdev_t *vd);
extern void vdev_stat_update(zio_t *zio, uint64_t psize);
Expand Down
2 changes: 2 additions & 0 deletions include/sys/vdev_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -280,6 +280,7 @@ struct vdev {
uint64_t vdev_ms_shift; /* metaslab size shift */
uint64_t vdev_ms_count; /* number of metaslabs */
metaslab_group_t *vdev_mg; /* metaslab group */
metaslab_group_t *vdev_log_mg; /* embedded slog metaslab group */
metaslab_t **vdev_ms; /* metaslab array */
uint64_t vdev_pending_fastwrite; /* allocated fastwrites */
txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */
Expand Down Expand Up @@ -636,6 +637,7 @@ extern int vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise);
* Other miscellaneous functions
*/
int vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj);
void vdev_metaslab_group_create(vdev_t *vd);

/*
* Vdev ashift optimization tunables
Expand Down
1 change: 1 addition & 0 deletions include/sys/zfs_debug.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,6 +56,7 @@ extern int zfs_dbgmsg_enable;
#define ZFS_DEBUG_INDIRECT_REMAP (1 << 10)
#define ZFS_DEBUG_TRIM (1 << 11)
#define ZFS_DEBUG_LOG_SPACEMAP (1 << 12)
#define ZFS_DEBUG_METASLAB_ALLOC (1 << 13)

extern void __set_error(const char *file, const char *func, int line, int err);
extern void __zfs_dbgmsg(char *buf);
Expand Down
16 changes: 16 additions & 0 deletions man/man5/zfs-module-parameters.5
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3936,6 +3936,22 @@ to limit potential SLOG device abuse by single active ZIL writer.
Default value: \fB786,432\fR.
.RE

.sp
.ne 2
.na
\fBzfs_embedded_slog_min_ms\fR (int)
.ad
.RS 12n
Usually, one metaslab from each (normal-class) vdev is dedicated for use by
the ZIL (to log synchronous writes).
However, if there are fewer than zfs_embedded_slog_min_ms metaslabs in the
vdev, this functionality is disabled.
This ensures that we don't set aside an unreasonable amount of space for the
ZIL.
.sp
Default value: \fB64\fR.
.RE

.sp
.ne 2
.na
Expand Down
94 changes: 61 additions & 33 deletions module/zfs/metaslab.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -524,7 +524,7 @@ metaslab_class_histogram_verify(metaslab_class_t *mc)

for (int c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
metaslab_group_t *mg = tvd->vdev_mg;
metaslab_group_t *mg = vdev_get_mg(tvd, mc);

/*
* Skip any holes, uninitialized top-levels, or
Expand All @@ -535,12 +535,16 @@ metaslab_class_histogram_verify(metaslab_class_t *mc)
continue;
}

IMPLY(mg == mg->mg_vd->vdev_log_mg,
mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));

for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++)
mc_hist[i] += mg->mg_histogram[i];
}

for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++)
for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
VERIFY3U(mc_hist[i], ==, mc->mc_histogram[i]);
}

kmem_free(mc_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
}
Expand Down Expand Up @@ -1004,16 +1008,22 @@ metaslab_group_initialized(metaslab_group_t *mg)
uint64_t
metaslab_group_get_space(metaslab_group_t *mg)
{
return ((1ULL << mg->mg_vd->vdev_ms_shift) * mg->mg_vd->vdev_ms_count);
/*
* Note that the number of nodes in mg_metaslab_tree may be one less
* than vdev_ms_count, due to the embedded log metaslab.
*/
mutex_enter(&mg->mg_lock);
uint64_t ms_count = avl_numnodes(&mg->mg_metaslab_tree);
mutex_exit(&mg->mg_lock);
return ((1ULL << mg->mg_vd->vdev_ms_shift) * ms_count);
}

void
metaslab_group_histogram_verify(metaslab_group_t *mg)
{
uint64_t *mg_hist;
vdev_t *vd = mg->mg_vd;
uint64_t ashift = vd->vdev_ashift;
int i;
avl_tree_t *t = &mg->mg_metaslab_tree;
uint64_t ashift = mg->mg_vd->vdev_ashift;

if ((zfs_flags & ZFS_DEBUG_HISTOGRAM_VERIFY) == 0)
return;
Expand All @@ -1024,21 +1034,25 @@ metaslab_group_histogram_verify(metaslab_group_t *mg)
ASSERT3U(RANGE_TREE_HISTOGRAM_SIZE, >=,
SPACE_MAP_HISTOGRAM_SIZE + ashift);

for (int m = 0; m < vd->vdev_ms_count; m++) {
metaslab_t *msp = vd->vdev_ms[m];

/* skip if not active or not a member */
if (msp->ms_sm == NULL || msp->ms_group != mg)
mutex_enter(&mg->mg_lock);
for (metaslab_t *msp = avl_first(t);
msp != NULL; msp = AVL_NEXT(t, msp)) {
VERIFY3P(msp->ms_group, ==, mg);
/* skip if not active */
if (msp->ms_sm == NULL)
continue;

for (i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++)
for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
mg_hist[i + ashift] +=
msp->ms_sm->sm_phys->smp_histogram[i];
}
}

for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i ++)
for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i ++)
VERIFY3U(mg_hist[i], ==, mg->mg_histogram[i]);

mutex_exit(&mg->mg_lock);
ahrens marked this conversation as resolved.
Show resolved Hide resolved

kmem_free(mg_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
}

Expand All @@ -1054,6 +1068,8 @@ metaslab_group_histogram_add(metaslab_group_t *mg, metaslab_t *msp)

mutex_enter(&mg->mg_lock);
for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
IMPLY(mg == mg->mg_vd->vdev_log_mg,
mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));
mg->mg_histogram[i + ashift] +=
msp->ms_sm->sm_phys->smp_histogram[i];
mc->mc_histogram[i + ashift] +=
Expand All @@ -1078,6 +1094,8 @@ metaslab_group_histogram_remove(metaslab_group_t *mg, metaslab_t *msp)
msp->ms_sm->sm_phys->smp_histogram[i]);
ASSERT3U(mc->mc_histogram[i + ashift], >=,
msp->ms_sm->sm_phys->smp_histogram[i]);
IMPLY(mg == mg->mg_vd->vdev_log_mg,
mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));

mg->mg_histogram[i + ashift] -=
msp->ms_sm->sm_phys->smp_histogram[i];
Expand Down Expand Up @@ -2741,37 +2759,47 @@ metaslab_fini(metaslab_t *msp)

mutex_enter(&msp->ms_lock);
VERIFY(msp->ms_group == NULL);
metaslab_space_update(vd, mg->mg_class,
-metaslab_allocated_space(msp), 0, -msp->ms_size);
/*
* If the range trees haven't been allocated, this metaslab hasn't
* been through metaslab_sync_done() for the first time yet, so its
* space hasn't been accounted for in its vdev and doesn't need to be
* subtracted.
*/
if (msp->ms_freed != NULL) {
metaslab_space_update(vd, mg->mg_class,
-metaslab_allocated_space(msp), 0, -msp->ms_size);

}
space_map_close(msp->ms_sm);
msp->ms_sm = NULL;

metaslab_unload(msp);

range_tree_destroy(msp->ms_allocatable);
range_tree_destroy(msp->ms_freeing);
range_tree_destroy(msp->ms_freed);

ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
metaslab_unflushed_changes_memused(msp));
spa->spa_unflushed_stats.sus_memused -=
metaslab_unflushed_changes_memused(msp);
range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
range_tree_destroy(msp->ms_unflushed_allocs);
range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);
range_tree_destroy(msp->ms_unflushed_frees);
if (msp->ms_freed != NULL) {
range_tree_destroy(msp->ms_freeing);
range_tree_destroy(msp->ms_freed);

for (int t = 0; t < TXG_SIZE; t++) {
range_tree_destroy(msp->ms_allocating[t]);
}
ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
metaslab_unflushed_changes_memused(msp));
spa->spa_unflushed_stats.sus_memused -=
metaslab_unflushed_changes_memused(msp);
range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
range_tree_destroy(msp->ms_unflushed_allocs);
range_tree_destroy(msp->ms_checkpointing);
range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);
range_tree_destroy(msp->ms_unflushed_frees);

for (int t = 0; t < TXG_DEFER_SIZE; t++) {
range_tree_destroy(msp->ms_defer[t]);
for (int t = 0; t < TXG_SIZE; t++) {
range_tree_destroy(msp->ms_allocating[t]);
}
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
range_tree_destroy(msp->ms_defer[t]);
}
}
ASSERT0(msp->ms_deferspace);

range_tree_destroy(msp->ms_checkpointing);

for (int t = 0; t < TXG_SIZE; t++)
ASSERT(!txg_list_member(&vd->vdev_ms_list, msp, t));

Expand Down Expand Up @@ -5113,7 +5141,7 @@ metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
* all else fails.
*/
if (vd != NULL && vd->vdev_mg != NULL) {
mg = vd->vdev_mg;
mg = vdev_get_mg(vd, mc);

if (flags & METASLAB_HINTBP_AVOID &&
mg->mg_next != NULL)
Expand Down
Loading