c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2024-12-26 03:09:34 +03:00
Code Issues Packages Projects Releases Wiki Activity

Add request size histograms (-r) to zpool iostat, minor man page fix

Browse Source 
Add -r option to "zpool iostat" to print request size histograms for the leaf
ZIOs. This includes histograms of individual ZIOs ("ind") and aggregate ZIOs
("agg"). These stats can be useful for seeing how well the ZFS IO aggregator
is working.

$ zpool iostat -r
mypool        sync_read    sync_write    async_read    async_write      scrub
req_size      ind    agg    ind    agg    ind    agg    ind    agg    ind    agg
----------  -----  -----  -----  -----  -----  -----  -----  -----  -----  -----
512             0      0      0      0      0      0    530      0      0      0
1K              0      0    260      0      0      0    116    246      0      0
2K              0      0      0      0      0      0      0    431      0      0
4K              0      0      0      0      0      0      3    107      0      0
8K             15      0     35      0      0      0      0      6      0      0
16K             0      0      0      0      0      0      0     39      0      0
32K             0      0      0      0      0      0      0      0      0      0
64K            20      0     40      0      0      0      0      0      0      0
128K            0      0     20      0      0      0      0      0      0      0
256K            0      0      0      0      0      0      0      0      0      0
512K            0      0      0      0      0      0      0      0      0      0
1M              0      0      0      0      0      0      0      0      0      0
2M              0      0      0      0      0      0      0      0      0      0
4M              0      0      0      0      0      0    155     19      0      0
8M              0      0      0      0      0      0      0    811      0      0
16M             0      0      0      0      0      0      0     68      0      0
--------------------------------------------------------------------------------

Also rename the stray "-G" in the man page to be "-w" for latency histograms.

Signed-off-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tim Chase <tim@chase2k.com>
Closes #4659
This commit is contained in:
Tony Hutter 2016-05-25 14:21:35 -07:00 committed by Brian Behlendorf
parent 4442f60d8e
commit 7e945072d1
7 changed files with 301 additions and 106 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

272
cmd/zpool/zpool_main.c
View File

@ -153,6 +153,7 @@ enum iostat_type {
IOS_LATENCY = 1, IOS_LATENCY = 1,
IOS_QUEUES = 2, IOS_QUEUES = 2,
IOS_L_HISTO = 3, IOS_L_HISTO = 3,
IOS_RQ_HISTO = 4,
IOS_COUNT, /* always last element */ IOS_COUNT, /* always last element */
}; };
@ -161,6 +162,62 @@ enum iostat_type {
#define IOS_LATENCY_M (1ULL << IOS_LATENCY) #define IOS_LATENCY_M (1ULL << IOS_LATENCY)
#define IOS_QUEUES_M (1ULL << IOS_QUEUES) #define IOS_QUEUES_M (1ULL << IOS_QUEUES)
#define IOS_L_HISTO_M (1ULL << IOS_L_HISTO) #define IOS_L_HISTO_M (1ULL << IOS_L_HISTO)
#define IOS_RQ_HISTO_M (1ULL << IOS_RQ_HISTO)
/* Mask of all the histo bits */
#define IOS_ANYHISTO_M (IOS_L_HISTO_M | IOS_RQ_HISTO_M)
/*
* Lookup table for iostat flags to nvlist names. Basically a list
* of all the nvlists a flag requires. Also specifies the order in
* which data gets printed in zpool iostat.
*/
static const char *vsx_type_to_nvlist[IOS_COUNT][11] = {
[IOS_L_HISTO] = {
ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
NULL},
[IOS_LATENCY] = {
ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
NULL},
[IOS_QUEUES] = {
ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
NULL},
[IOS_RQ_HISTO] = {
ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO,
ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO,
ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO,
ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO,
ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO,
ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO,
NULL},
};
/*
* Given a cb->cb_flags with a histogram bit set, return the iostat_type.
* Right now, only one histo bit is ever set at one time, so we can
* just do a highbit64(a)
*/
#define IOS_HISTO_IDX(a) (highbit64(a & IOS_ANYHISTO_M) - 1)
typedef struct zpool_command { typedef struct zpool_command {
const char *name; const char *name;
@ -255,7 +312,8 @@ get_usage(zpool_help_t idx) {
"[-R root] [-F [-n]]\n" "[-R root] [-F [-n]]\n"
"\t <pool | id> [newpool]\n")); "\t <pool | id> [newpool]\n"));
case HELP_IOSTAT: case HELP_IOSTAT:
return (gettext("\tiostat [-T d | u] [-ghHLpPvy] [[-lq]|-w]\n" return (gettext("\tiostat [-T d | u] [-ghHLpPvy] "
"[[-lq]|[-r|-w]]\n"
"\t [[pool ...]|[pool vdev ...]|[vdev ...]] " "\t [[pool ...]|[pool vdev ...]|[vdev ...]] "
"[interval [count]]\n")); "[interval [count]]\n"));
case HELP_LABELCLEAR: case HELP_LABELCLEAR:
@ -2531,6 +2589,9 @@ static const name_and_columns_t iostat_top_labels[][IOSTAT_MAX_LABELS] =
{NULL}}, {NULL}},
[IOS_L_HISTO] = {{"total_wait", 2}, {"disk_wait", 2}, [IOS_L_HISTO] = {{"total_wait", 2}, {"disk_wait", 2},
{"sync_queue", 2}, {"async_queue", 2}, {NULL}}, {"sync_queue", 2}, {"async_queue", 2}, {NULL}},
[IOS_RQ_HISTO] = {{"sync_read", 2}, {"sync_write", 2},
{"async_read", 2}, {"async_write", 2}, {"scrub", 2}, {NULL}},
}; };
/* Shorthand - if "columns" field not set, default to 1 column */ /* Shorthand - if "columns" field not set, default to 1 column */
@ -2544,6 +2605,13 @@ static const name_and_columns_t iostat_bottom_labels[][IOSTAT_MAX_LABELS] =
{"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"}, {NULL}}, {"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"}, {NULL}},
[IOS_L_HISTO] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"}, [IOS_L_HISTO] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"},
{"write"}, {"read"}, {"write"}, {"scrub"}, {NULL}}, {"write"}, {"read"}, {"write"}, {"scrub"}, {NULL}},
[IOS_RQ_HISTO] = {{"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"},
{"ind"}, {"agg"}, {"ind"}, {"agg"}, {NULL}},
};
static const char *histo_to_title[] = {
[IOS_L_HISTO] = "latency",
[IOS_RQ_HISTO] = "req_size",
}; };
/* /*
@ -2562,6 +2630,25 @@ label_array_len(const name_and_columns_t *labels)
return (i); return (i);
} }
/*
* Return the number of strings in a null-terminated string array.
* For example:
*
* const char foo[] = {"bar", "baz", NULL}
*
* returns 2
*/
static uint64_t
str_array_len(const char *array[])
{
uint64_t i = 0;
while (array[i])
i++;
return (i);
}
/* /*
* Return a default column width for default/latency/queue columns. This does * Return a default column width for default/latency/queue columns. This does
* not include histograms, which have their columns autosized. * not include histograms, which have their columns autosized.
@ -2673,14 +2760,22 @@ print_iostat_dashes(iostat_cbdata_t *cb, unsigned int force_column_width,
uint64_t f; uint64_t f;
int idx; int idx;
const name_and_columns_t *labels; const name_and_columns_t *labels;
const char *title;
if (cb->cb_flags & IOS_ANYHISTO_M) {
title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
} else if (cb->cb_vdev_names_count) {
title = "vdev";
} else {
title = "pool";
}
namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
name ? strlen(name) : 0);
if (cb->cb_flags & IOS_L_HISTO_M)
namewidth = MAX(cb->cb_namewidth, strlen("latency"));
else
namewidth = cb->cb_namewidth;
if (name) { if (name) {
namewidth = MAX(cb->cb_namewidth, strlen(name));
printf("%-*s", namewidth, name); printf("%-*s", namewidth, name);
} else { } else {
for (i = 0; i < namewidth; i++) for (i = 0; i < namewidth; i++)
@ -2727,22 +2822,28 @@ print_iostat_header_impl(iostat_cbdata_t *cb, unsigned int force_column_width,
const char *histo_vdev_name) const char *histo_vdev_name)
{ {
unsigned int namewidth; unsigned int namewidth;
uint64_t flags = cb->cb_flags; const char *title;
if (flags & IOS_L_HISTO_M) if (cb->cb_flags & IOS_ANYHISTO_M) {
namewidth = MAX(cb->cb_namewidth, strlen("latency")); title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
else } else if (cb->cb_vdev_names_count) {
namewidth = cb->cb_namewidth; title = "vdev";
} else {
title = "pool";
}
if (flags & IOS_L_HISTO_M) namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
histo_vdev_name ? strlen(histo_vdev_name) : 0);
if (histo_vdev_name)
printf("%-*s", namewidth, histo_vdev_name); printf("%-*s", namewidth, histo_vdev_name);
else else
printf("%*s", namewidth, ""); printf("%*s", namewidth, "");
print_iostat_labels(cb, force_column_width, iostat_top_labels); print_iostat_labels(cb, force_column_width, iostat_top_labels);
printf("%-*s", namewidth, flags & IOS_L_HISTO_M ? "latency" : printf("%-*s", namewidth, title);
cb->cb_vdev_names_count ? "vdev" : "pool");
print_iostat_labels(cb, force_column_width, iostat_bottom_labels); print_iostat_labels(cb, force_column_width, iostat_bottom_labels);
@ -2918,6 +3019,7 @@ print_iostat_histo(struct stat_array *nva, unsigned int len,
uint64_t val; uint64_t val;
enum zfs_nicenum_format format; enum zfs_nicenum_format format;
unsigned int buckets; unsigned int buckets;
unsigned int start_bucket;
if (cb->cb_literal) if (cb->cb_literal)
format = ZFS_NICENUM_RAW; format = ZFS_NICENUM_RAW;
@ -2927,12 +3029,25 @@ print_iostat_histo(struct stat_array *nva, unsigned int len,
/* All these histos are the same size, so just use nva[0].count */ /* All these histos are the same size, so just use nva[0].count */
buckets = nva[0].count; buckets = nva[0].count;
for (j = 0; j < buckets; j++) { if (cb->cb_flags & IOS_RQ_HISTO_M) {
/* Start at 512 - req size should never be lower than this */
start_bucket = 9;
} else {
start_bucket = 0;
}
for (j = start_bucket; j < buckets; j++) {
/* Print histogram bucket label */
if (cb->cb_flags & IOS_L_HISTO_M) {
/* Ending range of this bucket */ /* Ending range of this bucket */
val = (1UL << (j + 1)) - 1; val = (1UL << (j + 1)) - 1;
/* Print histogram bucket label */
zfs_nicetime(val, buf, sizeof (buf)); zfs_nicetime(val, buf, sizeof (buf));
} else {
/* Request size (starting range of bucket) */
val = (1UL << j);
zfs_nicenum(val, buf, sizeof (buf));
}
if (cb->cb_scripted) if (cb->cb_scripted)
printf("%llu", (u_longlong_t) val); printf("%llu", (u_longlong_t) val);
else else
@ -2962,30 +3077,29 @@ print_iostat_histos(iostat_cbdata_t *cb, nvlist_t *oldnv,
unsigned int column_width; unsigned int column_width;
unsigned int namewidth; unsigned int namewidth;
unsigned int entire_width; unsigned int entire_width;
enum iostat_type type;
const char *names[] = {
ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
};
struct stat_array *nva; struct stat_array *nva;
nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), oldnv, newnv); const char **names;
unsigned int names_len;
/* What type of histo are we? */
type = IOS_HISTO_IDX(cb->cb_flags);
/* Get NULL-terminated array of nvlist names for our histo */
names = vsx_type_to_nvlist[type];
names_len = str_array_len(names); /* num of names */
nva = calc_and_alloc_stats_ex(names, names_len, oldnv, newnv);
if (cb->cb_literal) { if (cb->cb_literal) {
column_width = MAX(5, column_width = MAX(5,
(unsigned int) log10(stat_histo_max(nva, (unsigned int) log10(stat_histo_max(nva, names_len)) + 1);
ARRAY_SIZE(names))) + 1);
} else { } else {
column_width = 5; column_width = 5;
} }
namewidth = MAX(cb->cb_namewidth, strlen("latency")); namewidth = MAX(cb->cb_namewidth,
strlen(histo_to_title[IOS_HISTO_IDX(cb->cb_flags)]));
/* /*
* Calculate the entire line width of what we're printing. The * Calculate the entire line width of what we're printing. The
@ -2998,17 +3112,17 @@ print_iostat_histos(iostat_cbdata_t *cb, nvlist_t *oldnv,
/* |__________| <--- entire_width */ /* |__________| <--- entire_width */
/* */ /* */
entire_width = namewidth + (column_width + 2) * entire_width = namewidth + (column_width + 2) *
label_array_len(iostat_bottom_labels[IOS_L_HISTO]); label_array_len(iostat_bottom_labels[type]);
if (cb->cb_scripted) if (cb->cb_scripted)
printf("%s\n", name); printf("%s\n", name);
else else
print_iostat_header_impl(cb, column_width, name); print_iostat_header_impl(cb, column_width, name);
print_iostat_histo(nva, ARRAY_SIZE(names), cb, column_width, print_iostat_histo(nva, names_len, cb, column_width,
namewidth, scale); namewidth, scale);
free_calc_stats(nva, ARRAY_SIZE(names)); free_calc_stats(nva, names_len);
if (!cb->cb_scripted) if (!cb->cb_scripted)
print_solid_separator(entire_width); print_solid_separator(entire_width);
} }
@ -3219,7 +3333,7 @@ print_vdev_stats(zpool_handle_t *zhp, const char *name, nvlist_t *oldnv,
* Print the vdev name unless it's is a histogram. Histograms * Print the vdev name unless it's is a histogram. Histograms
* display the vdev name in the header itself. * display the vdev name in the header itself.
*/ */
if (!(cb->cb_flags & IOS_L_HISTO_M)) { if (!(cb->cb_flags & IOS_ANYHISTO_M)) {
if (cb->cb_scripted) { if (cb->cb_scripted) {
printf("%s", name); printf("%s", name);
} else { } else {
@ -3234,7 +3348,7 @@ print_vdev_stats(zpool_handle_t *zhp, const char *name, nvlist_t *oldnv,
/* Calculate our scaling factor */ /* Calculate our scaling factor */
tdelta = newvs->vs_timestamp - oldvs->vs_timestamp; tdelta = newvs->vs_timestamp - oldvs->vs_timestamp;
if ((oldvs->vs_timestamp == 0) && (cb->cb_flags & IOS_L_HISTO_M)) { if ((oldvs->vs_timestamp == 0) && (cb->cb_flags & IOS_ANYHISTO_M)) {
/* /*
* If we specify printing histograms with no time interval, then * If we specify printing histograms with no time interval, then
* print the histogram numbers over the entire lifetime of the * print the histogram numbers over the entire lifetime of the
@ -3256,12 +3370,12 @@ print_vdev_stats(zpool_handle_t *zhp, const char *name, nvlist_t *oldnv,
print_iostat_latency(cb, oldnv, newnv, scale); print_iostat_latency(cb, oldnv, newnv, scale);
if (cb->cb_flags & IOS_QUEUES_M) if (cb->cb_flags & IOS_QUEUES_M)
print_iostat_queues(cb, oldnv, newnv, scale); print_iostat_queues(cb, oldnv, newnv, scale);
if (cb->cb_flags & IOS_L_HISTO_M) { if (cb->cb_flags & IOS_ANYHISTO_M) {
printf("\n"); printf("\n");
print_iostat_histos(cb, oldnv, newnv, scale, name); print_iostat_histos(cb, oldnv, newnv, scale, name);
} }
if (!(cb->cb_flags & IOS_L_HISTO_M)) if (!(cb->cb_flags & IOS_ANYHISTO_M))
printf("\n"); printf("\n");
free(calcvs); free(calcvs);
@ -3303,7 +3417,7 @@ children:
*/ */
if (num_logs(newnv) > 0) { if (num_logs(newnv) > 0) {
if ((!(cb->cb_flags & IOS_L_HISTO_M)) && !cb->cb_scripted && if ((!(cb->cb_flags & IOS_ANYHISTO_M)) && !cb->cb_scripted &&
!cb->cb_vdev_names) { !cb->cb_vdev_names) {
print_iostat_dashes(cb, 0, "logs"); print_iostat_dashes(cb, 0, "logs");
} }
@ -3337,7 +3451,7 @@ children:
return (ret); return (ret);
if (children > 0) { if (children > 0) {
if ((!(cb->cb_flags & IOS_L_HISTO_M)) && !cb->cb_scripted && if ((!(cb->cb_flags & IOS_ANYHISTO_M)) && !cb->cb_scripted &&
!cb->cb_vdev_names) { !cb->cb_vdev_names) {
print_iostat_dashes(cb, 0, "cache"); print_iostat_dashes(cb, 0, "cache");
} }
@ -3399,9 +3513,10 @@ print_iostat(zpool_handle_t *zhp, void *data)
ret = print_vdev_stats(zhp, zpool_get_name(zhp), oldnvroot, newnvroot, ret = print_vdev_stats(zhp, zpool_get_name(zhp), oldnvroot, newnvroot,
cb, 0); cb, 0);
if ((ret != 0) && !(cb->cb_flags & IOS_L_HISTO_M) && !cb->cb_scripted && if ((ret != 0) && !(cb->cb_flags & IOS_ANYHISTO_M) &&
cb->cb_verbose && !cb->cb_vdev_names_count) !cb->cb_scripted && cb->cb_verbose && !cb->cb_vdev_names_count) {
print_iostat_separator(cb); print_iostat_separator(cb);
}
return (ret); return (ret);
} }
@ -3553,37 +3668,6 @@ get_stat_flags_cb(zpool_handle_t *zhp, void *data)
uint64_t flags = 0; uint64_t flags = 0;
int i, j; int i, j;
/*
* Lookup table for extended iostat flags to nvlist names.
* Basically a list of all the nvpairs a flag requires.
*/
static const char *vsx_type_to_nvlist[IOS_COUNT][10] = {
[IOS_L_HISTO] = {
ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
NULL},
[IOS_LATENCY] = {
ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
NULL},
[IOS_QUEUES] = {
ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
NULL}
};
config = zpool_get_config(zhp, NULL); config = zpool_get_config(zhp, NULL);
verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0); &nvroot) == 0);
@ -3818,7 +3902,7 @@ fsleep(float sec) {
/* /*
* zpool iostat [-ghHLpPvy] [[-lq]-w] [-n name] [-T d|u] * zpool iostat [-ghHLpPvy] [[-lq]|[-r|-w]] [-n name] [-T d|u]
* [[ pool ...]|[pool vdev ...]|[vdev ...]] * [[ pool ...]|[pool vdev ...]|[vdev ...]]
* [interval [count]] * [interval [count]]
* *
@ -3832,7 +3916,8 @@ fsleep(float sec) {
* by a single tab. * by a single tab.
* -l Display average latency * -l Display average latency
* -q Display queue depths * -q Display queue depths
* -w Display histograms * -w Display latency histograms
* -r Display request size histogram
* -T Display a timestamp in date(1) or Unix format * -T Display a timestamp in date(1) or Unix format
* *
* This command can be tricky because we want to be able to deal with pool * This command can be tricky because we want to be able to deal with pool
@ -3851,7 +3936,7 @@ zpool_do_iostat(int argc, char **argv)
unsigned long count = 0; unsigned long count = 0;
zpool_list_t *list; zpool_list_t *list;
boolean_t verbose = B_FALSE; boolean_t verbose = B_FALSE;
boolean_t latency = B_FALSE, histo = B_FALSE; boolean_t latency = B_FALSE, l_histo = B_FALSE, rq_histo = B_FALSE;
boolean_t queues = B_FALSE, parsable = B_FALSE, scripted = B_FALSE; boolean_t queues = B_FALSE, parsable = B_FALSE, scripted = B_FALSE;
boolean_t omit_since_boot = B_FALSE; boolean_t omit_since_boot = B_FALSE;
boolean_t guid = B_FALSE; boolean_t guid = B_FALSE;
@ -3861,12 +3946,12 @@ zpool_do_iostat(int argc, char **argv)
/* Used for printing error message */ /* Used for printing error message */
const char flag_to_arg[] = {[IOS_LATENCY] = 'l', [IOS_QUEUES] = 'q', const char flag_to_arg[] = {[IOS_LATENCY] = 'l', [IOS_QUEUES] = 'q',
[IOS_L_HISTO] = 'w'}; [IOS_L_HISTO] = 'w', [IOS_RQ_HISTO] = 'r'};
uint64_t unsupported_flags; uint64_t unsupported_flags;
/* check options */ /* check options */
while ((c = getopt(argc, argv, "gLPT:vyhplqwH")) != -1) { while ((c = getopt(argc, argv, "gLPT:vyhplqrwH")) != -1) {
switch (c) { switch (c) {
case 'g': case 'g':
guid = B_TRUE; guid = B_TRUE;
@ -3896,7 +3981,10 @@ zpool_do_iostat(int argc, char **argv)
scripted = B_TRUE; scripted = B_TRUE;
break; break;
case 'w': case 'w':
histo = B_TRUE; l_histo = B_TRUE;
break;
case 'r':
rq_histo = B_TRUE;
break; break;
case 'y': case 'y':
omit_since_boot = B_TRUE; omit_since_boot = B_TRUE;
@ -3997,10 +4085,18 @@ zpool_do_iostat(int argc, char **argv)
return (1); return (1);
} }
if (histo && (queues || latency)) { if ((l_histo || rq_histo) && (queues || latency)) {
pool_list_free(list); pool_list_free(list);
(void) fprintf(stderr, (void) fprintf(stderr,
gettext("-w isn't allowed with [-q|-l]\n")); gettext("[-r|-w] isn't allowed with [-q|-l]\n"));
usage(B_FALSE);
return (1);
}
if (l_histo && rq_histo) {
pool_list_free(list);
(void) fprintf(stderr,
gettext("Only one of [-r|-w] can be passed at a time\n"));
usage(B_FALSE); usage(B_FALSE);
return (1); return (1);
} }
@ -4010,13 +4106,15 @@ zpool_do_iostat(int argc, char **argv)
*/ */
cb.cb_list = list; cb.cb_list = list;
if (histo) { if (l_histo) {
/* /*
* Histograms tables look out of place when you try to display * Histograms tables look out of place when you try to display
* them with the other stats, so make a rule that you can only * them with the other stats, so make a rule that you can only
* print histograms by themselves. * print histograms by themselves.
*/ */
cb.cb_flags = IOS_L_HISTO_M; cb.cb_flags = IOS_L_HISTO_M;
} else if (rq_histo) {
cb.cb_flags = IOS_RQ_HISTO_M;
} else { } else {
cb.cb_flags = IOS_DEFAULT_M; cb.cb_flags = IOS_DEFAULT_M;
if (latency) if (latency)
@ -4088,7 +4186,7 @@ zpool_do_iostat(int argc, char **argv)
*/ */
if (((++cb.cb_iteration == 1 && !skip) || if (((++cb.cb_iteration == 1 && !skip) ||
(skip != verbose)) && (skip != verbose)) &&
(!(cb.cb_flags & IOS_L_HISTO_M)) && (!(cb.cb_flags & IOS_ANYHISTO_M)) &&
!cb.cb_scripted) !cb.cb_scripted)
print_iostat_header(&cb); print_iostat_header(&cb);
@ -4108,8 +4206,8 @@ zpool_do_iostat(int argc, char **argv)
* we also want an ending separator. * we also want an ending separator.
*/ */
if (((npools > 1 && !verbose && if (((npools > 1 && !verbose &&
!(cb.cb_flags & IOS_L_HISTO_M)) || !(cb.cb_flags & IOS_ANYHISTO_M)) ||
(!(cb.cb_flags & IOS_L_HISTO_M) && (!(cb.cb_flags & IOS_ANYHISTO_M) &&
cb.cb_vdev_names_count)) && cb.cb_vdev_names_count)) &&
!cb.cb_scripted) { !cb.cb_scripted) {
print_iostat_separator(&cb); print_iostat_separator(&cb);

37
include/sys/fs/zfs.h
View File

@ -558,7 +558,17 @@ typedef struct zpool_rewind_policy {
#define ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO "vdev_async_w_lat_histo" #define ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO "vdev_async_w_lat_histo"
#define ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO "vdev_scrub_histo" #define ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO "vdev_scrub_histo"
/* Request size histograms */
#define ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO "vdev_sync_ind_r_histo"
#define ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO "vdev_sync_ind_w_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO "vdev_async_ind_r_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO "vdev_async_ind_w_histo"
#define ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO "vdev_ind_scrub_histo"
#define ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO "vdev_sync_agg_r_histo"
#define ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO "vdev_sync_agg_w_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO "vdev_async_agg_r_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO "vdev_async_agg_w_histo"
#define ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO "vdev_agg_scrub_histo"
#define ZPOOL_CONFIG_WHOLE_DISK "whole_disk" #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk"
#define ZPOOL_CONFIG_ERRCOUNT "error_count" #define ZPOOL_CONFIG_ERRCOUNT "error_count"
@ -824,20 +834,33 @@ typedef struct vdev_stat_ex {
* 2^37 nanoseconds = 134s. Timeouts will probably start kicking in * 2^37 nanoseconds = 134s. Timeouts will probably start kicking in
* before this. * before this.
*/ */
#define VDEV_HISTO_BUCKETS 37 #define VDEV_L_HISTO_BUCKETS 37 /* Latency histo buckets */
#define VDEV_RQ_HISTO_BUCKETS 25 /* Request size histo buckets */
/* Amount of time in ZIO queue (ns) */ /* Amount of time in ZIO queue (ns) */
uint64_t vsx_queue_histo[ZIO_PRIORITY_NUM_QUEUEABLE] uint64_t vsx_queue_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
[VDEV_HISTO_BUCKETS]; [VDEV_L_HISTO_BUCKETS];
/* Total ZIO latency (ns). Includes queuing and disk access time */ /* Total ZIO latency (ns). Includes queuing and disk access time */
uint64_t vsx_total_histo[ZIO_TYPES][VDEV_HISTO_BUCKETS]; uint64_t vsx_total_histo[ZIO_TYPES][VDEV_L_HISTO_BUCKETS];
/* Amount of time to read/write the disk (ns) */ /* Amount of time to read/write the disk (ns) */
uint64_t vsx_disk_histo[ZIO_TYPES][VDEV_HISTO_BUCKETS]; uint64_t vsx_disk_histo[ZIO_TYPES][VDEV_L_HISTO_BUCKETS];
/* "lookup the bucket for a value" macro */ /* "lookup the bucket for a value" histogram macros */
#define HISTO(a) (a != 0 ? MIN(highbit64(a) - 1, VDEV_HISTO_BUCKETS - 1) : 0) #define HISTO(val, buckets) (val != 0 ? MIN(highbit64(val) - 1, \
buckets - 1) : 0)
#define L_HISTO(a) HISTO(a, VDEV_L_HISTO_BUCKETS)
#define RQ_HISTO(a) HISTO(a, VDEV_RQ_HISTO_BUCKETS)
/* Physical IO histogram */
uint64_t vsx_ind_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
[VDEV_RQ_HISTO_BUCKETS];
/* Delegated (aggregated) physical IO histogram */
uint64_t vsx_agg_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
[VDEV_RQ_HISTO_BUCKETS];
} vdev_stat_ex_t; } vdev_stat_ex_t;

17
man/man8/zpool.8
View File

@ -95,7 +95,7 @@ zpool \- configures ZFS storage pools
.LP .LP
.nf .nf
\fB\fBzpool iostat\fR [\fB-T\fR \fBd\fR | \fBu\fR] [\fB-ghHLpPvy\fR] [\fB-G\fR|[\fB-lq\fR]] \fB\fBzpool iostat\fR [\fB-T\fR \fBd\fR | \fBu\fR] [\fB-ghHLpPvy\fR] [\fB-lq\fR]|[\fB-r\fR|-\fBw\fR]]
[[\fIpool\fR ...]|[\fIpool vdev\fR ...]|[\fIvdev\fR ...]] [\fIinterval\fR[\fIcount\fR]]\fR [[\fIpool\fR ...]|[\fIpool vdev\fR ...]|[\fIvdev\fR ...]] [\fIinterval\fR[\fIcount\fR]]\fR
.fi .fi
@ -1510,7 +1510,7 @@ Scan using the default search path, the libblkid cache will not be consulted. A
.sp .sp
.ne 2 .ne 2
.na .na
\fB\fBzpool iostat\fR [\fB-T\fR \fBd\fR | \fBu\fR] [\fB-ghHLpPvy\fR] [\fB-w\fR|[\fB-lq\fR]] [[\fIpool\fR ...]|[\fIpool vdev\fR ...]|[\fIvdev\fR ...]] [\fIinterval\fR[\fIcount\fR]]\fR \fB\fBzpool iostat\fR [\fB-T\fR \fBd\fR | \fBu\fR] [\fB-ghHLpPvy\fR] [[\fB-lq\fR]|[\fB-r\fR|\fB-w\fR]] [[\fIpool\fR ...]|[\fIpool vdev\fR ...]|[\fIvdev\fR ...]] [\fIinterval\fR[\fIcount\fR]]\fR
.ad .ad
.sp .6 .sp .6
@ -1582,6 +1582,19 @@ Display numbers in parsable (exact) values. Time values are in nanoseconds.
Display full paths for vdevs instead of only the last component of the path. This can be used in conjunction with the \fB-L\fR flag. Display full paths for vdevs instead of only the last component of the path. This can be used in conjunction with the \fB-L\fR flag.
.RE .RE
.sp
.ne 2
.na
\fB\fB-r\fR\fR
.ad
.RS 12n
Print request size histograms for the leaf ZIOs. This includes histograms of
individual ZIOs ("ind") and aggregate ZIOs ("agg"). These stats can be useful
for seeing how well the ZFS IO aggregator is working. Do not confuse these
request size stats with the block layer requests; it's possible ZIOs can
be broken up before being sent to the block device.
.RE
.sp .sp
.ne 2 .ne 2
.na .na

27
module/zfs/vdev.c
View File

@ -2784,21 +2784,30 @@ vdev_get_child_stat_ex(vdev_t *cvd, vdev_stat_ex_t *vsx, vdev_stat_ex_t *cvsx)
{ {
int t, b; int t, b;
for (t = 0; t < ZIO_TYPES; t++) { for (t = 0; t < ZIO_TYPES; t++) {
for (b = 0; b < VDEV_HISTO_BUCKETS; b++) { for (b = 0; b < ARRAY_SIZE(vsx->vsx_disk_histo[0]); b++)
vsx->vsx_disk_histo[t][b] += cvsx->vsx_disk_histo[t][b]; vsx->vsx_disk_histo[t][b] += cvsx->vsx_disk_histo[t][b];
for (b = 0; b < ARRAY_SIZE(vsx->vsx_total_histo[0]); b++) {
vsx->vsx_total_histo[t][b] += vsx->vsx_total_histo[t][b] +=
cvsx->vsx_total_histo[t][b]; cvsx->vsx_total_histo[t][b];
} }
} }
for (t = 0; t < ZIO_PRIORITY_NUM_QUEUEABLE; t++) { for (t = 0; t < ZIO_PRIORITY_NUM_QUEUEABLE; t++) {
for (b = 0; b < VDEV_HISTO_BUCKETS; b++) { for (b = 0; b < ARRAY_SIZE(vsx->vsx_queue_histo[0]); b++) {
vsx->vsx_queue_histo[t][b] += vsx->vsx_queue_histo[t][b] +=
cvsx->vsx_queue_histo[t][b]; cvsx->vsx_queue_histo[t][b];
} }
vsx->vsx_active_queue[t] += cvsx->vsx_active_queue[t]; vsx->vsx_active_queue[t] += cvsx->vsx_active_queue[t];
vsx->vsx_pend_queue[t] += cvsx->vsx_pend_queue[t]; vsx->vsx_pend_queue[t] += cvsx->vsx_pend_queue[t];
for (b = 0; b < ARRAY_SIZE(vsx->vsx_ind_histo[0]); b++)
vsx->vsx_ind_histo[t][b] += cvsx->vsx_ind_histo[t][b];
for (b = 0; b < ARRAY_SIZE(vsx->vsx_agg_histo[0]); b++)
vsx->vsx_agg_histo[t][b] += cvsx->vsx_agg_histo[t][b];
} }
} }
/* /*
@ -2974,13 +2983,21 @@ vdev_stat_update(zio_t *zio, uint64_t psize)
vs->vs_ops[type]++; vs->vs_ops[type]++;
vs->vs_bytes[type] += psize; vs->vs_bytes[type] += psize;
if (flags & ZIO_FLAG_DELEGATED) {
vsx->vsx_agg_histo[zio->io_priority]
[RQ_HISTO(zio->io_size)]++;
} else {
vsx->vsx_ind_histo[zio->io_priority]
[RQ_HISTO(zio->io_size)]++;
}
if (zio->io_delta && zio->io_delay) { if (zio->io_delta && zio->io_delay) {
vsx->vsx_queue_histo[zio->io_priority] vsx->vsx_queue_histo[zio->io_priority]
[HISTO(zio->io_delta - zio->io_delay)]++; [L_HISTO(zio->io_delta - zio->io_delay)]++;
vsx->vsx_disk_histo[type] vsx->vsx_disk_histo[type]
[HISTO(zio->io_delay)]++; [L_HISTO(zio->io_delay)]++;
vsx->vsx_total_histo[type] vsx->vsx_total_histo[type]
[HISTO(zio->io_delta)]++; [L_HISTO(zio->io_delta)]++;
} }
} }

41
module/zfs/vdev_label.c
View File

@ -302,6 +302,47 @@ vdev_config_generate_stats(vdev_t *vd, nvlist_t *nv)
vsx->vsx_queue_histo[ZIO_PRIORITY_SCRUB], vsx->vsx_queue_histo[ZIO_PRIORITY_SCRUB],
ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_SCRUB])); ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_SCRUB]));
/* Request sizes */
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO,
vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_READ],
ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_READ]));
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO,
vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_WRITE],
ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_WRITE]));
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO,
vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_READ],
ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_READ]));
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO,
vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_WRITE],
ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_WRITE]));
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO,
vsx->vsx_ind_histo[ZIO_PRIORITY_SCRUB],
ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_SCRUB]));
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO,
vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_READ],
ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_READ]));
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO,
vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_WRITE],
ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_WRITE]));
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO,
vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_READ],
ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_READ]));
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO,
vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_WRITE],
ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_WRITE]));
fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO,
vsx->vsx_agg_histo[ZIO_PRIORITY_SCRUB],
ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_SCRUB]));
/* Add extended stats nvlist to main nvlist */ /* Add extended stats nvlist to main nvlist */
fnvlist_add_nvlist(nv, ZPOOL_CONFIG_VDEV_STATS_EX, nvx); fnvlist_add_nvlist(nv, ZPOOL_CONFIG_VDEV_STATS_EX, nvx);

3
tests/zfs-tests/tests/functional/cli_user/zpool_iostat/zpool_iostat_003_neg.ksh
View File

@ -52,7 +52,8 @@ fi
set -A args "" "-?" "-f" "nonexistpool" "$TESTPOOL/$TESTFS" \ set -A args "" "-?" "-f" "nonexistpool" "$TESTPOOL/$TESTFS" \
"$testpool 0" "$testpool -1" "$testpool 1 0" \ "$testpool 0" "$testpool -1" "$testpool 1 0" \
"$testpool 0 0" "$testpool -wl" "$testpool -wq" "$testpool 0 0" "$testpool -wl" "$testpool -wq" "$testpool -wr" \
"$testpool -rq" "$testpool -lr"
log_assert "Executing 'zpool iostat' with bad options fails" log_assert "Executing 'zpool iostat' with bad options fails"

6
tests/zfs-tests/tests/functional/cli_user/zpool_iostat/zpool_iostat_004_pos.ksh
View File

@ -36,7 +36,7 @@
# #
# DESCRIPTION: # DESCRIPTION:
# Executing 'zpool iostat' command with various combinations of extended # Executing 'zpool iostat' command with various combinations of extended
# stats (-vqL), parsable/script options (-pH), and misc lists of pools # stats (-lqwr), parsable/script options (-pH), and misc lists of pools
# and vdevs. # and vdevs.
# #
# STRATEGY: # STRATEGY:
@ -59,7 +59,9 @@ set -A args "" "-v" "-q" "-l" "-lq $TESTPOOL" "-ql ${DISKS[0]} ${DISKS[1]}" \
"-wp $TESTPOOL" \ "-wp $TESTPOOL" \
"-qlH $TESTPOOL ${DISKS[0]}" \ "-qlH $TESTPOOL ${DISKS[0]}" \
"-vpH ${DISKS[0]}" \ "-vpH ${DISKS[0]}" \
"-wpH ${DISKS[0]}" "-wpH ${DISKS[0]}" \
"-r ${DISKS[0]}" \
"-rpH ${DISKS[0]}"
log_assert "Executing 'zpool iostat' with extended stat options succeeds" log_assert "Executing 'zpool iostat' with extended stat options succeeds"
log_note "testpool: $TESTPOOL, disks $DISKS" log_note "testpool: $TESTPOOL, disks $DISKS"