Archive-Team
/
zfs
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 103 Wiki Activity

zil: add stats for commit failure/fallback (#16315) 

There's no good way to tell when a ZIL commit fails and falls back to a
transaction sync, other than perhaps a throughput drop. This adds
counters so we can see when it happens and why.

Sponsored-by: Klara, Inc.
Sponsored-by: Wasabi Technology, Inc.

Signed-off-by: Rob Norris <rob.norris@klarasystems.com>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
This commit is contained in:
Rob Norris 2024-07-26 09:53:59 +10:00 committed by GitHub
parent 2fc646160f
commit 7ddc1f737f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 40 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
cmd/zilstat.in
View File

@ -43,6 +43,9 @@ cols = {
"obj": [12, -1, "objset"], "obj": [12, -1, "objset"],
"cc": [5, 1000, "zil_commit_count"], "cc": [5, 1000, "zil_commit_count"],
"cwc": [5, 1000, "zil_commit_writer_count"], "cwc": [5, 1000, "zil_commit_writer_count"],
"cec": [5, 1000, "zil_commit_error_count"],
"csc": [5, 1000, "zil_commit_stall_count"],
"cSc": [5, 1000, "zil_commit_suspend_count"],
"ic": [5, 1000, "zil_itx_count"], "ic": [5, 1000, "zil_itx_count"],
"iic": [5, 1000, "zil_itx_indirect_count"], "iic": [5, 1000, "zil_itx_indirect_count"],
"iib": [5, 1024, "zil_itx_indirect_bytes"], "iib": [5, 1024, "zil_itx_indirect_bytes"],

18
include/sys/zil.h
View File

@ -467,6 +467,21 @@ typedef struct zil_stats {
*/ */
kstat_named_t zil_commit_writer_count; kstat_named_t zil_commit_writer_count;
/*
* Number of times a ZIL commit failed and the ZIL was forced to fall
* back to txg_wait_synced(). The separate counts are for different
* reasons:
* - error: ZIL IO (write/flush) returned an error
* (see zil_commit_impl())
* - stall: LWB block allocation failed, ZIL chain abandoned
* (see zil_commit_writer_stall())
* - suspend: ZIL suspended
* (see zil_commit(), zil_get_commit_list())
*/
kstat_named_t zil_commit_error_count;
kstat_named_t zil_commit_stall_count;
kstat_named_t zil_commit_suspend_count;
/* /*
* Number of transactions (reads, writes, renames, etc.) * Number of transactions (reads, writes, renames, etc.)
* that have been committed. * that have been committed.
@ -510,6 +525,9 @@ typedef struct zil_stats {
typedef struct zil_sums { typedef struct zil_sums {
wmsum_t zil_commit_count; wmsum_t zil_commit_count;
wmsum_t zil_commit_writer_count; wmsum_t zil_commit_writer_count;
wmsum_t zil_commit_error_count;
wmsum_t zil_commit_stall_count;
wmsum_t zil_commit_suspend_count;
wmsum_t zil_itx_count; wmsum_t zil_itx_count;
wmsum_t zil_itx_indirect_count; wmsum_t zil_itx_indirect_count;
wmsum_t zil_itx_indirect_bytes; wmsum_t zil_itx_indirect_bytes;

19
module/zfs/zil.c
View File

@ -99,6 +99,9 @@ static uint_t zfs_commit_timeout_pct = 10;
static zil_kstat_values_t zil_stats = { static zil_kstat_values_t zil_stats = {
{ "zil_commit_count", KSTAT_DATA_UINT64 }, { "zil_commit_count", KSTAT_DATA_UINT64 },
{ "zil_commit_writer_count", KSTAT_DATA_UINT64 }, { "zil_commit_writer_count", KSTAT_DATA_UINT64 },
{ "zil_commit_error_count", KSTAT_DATA_UINT64 },
{ "zil_commit_stall_count", KSTAT_DATA_UINT64 },
{ "zil_commit_suspend_count", KSTAT_DATA_UINT64 },
{ "zil_itx_count", KSTAT_DATA_UINT64 }, { "zil_itx_count", KSTAT_DATA_UINT64 },
{ "zil_itx_indirect_count", KSTAT_DATA_UINT64 }, { "zil_itx_indirect_count", KSTAT_DATA_UINT64 },
{ "zil_itx_indirect_bytes", KSTAT_DATA_UINT64 }, { "zil_itx_indirect_bytes", KSTAT_DATA_UINT64 },
@ -360,6 +363,9 @@ zil_sums_init(zil_sums_t *zs)
{ {
wmsum_init(&zs->zil_commit_count, 0); wmsum_init(&zs->zil_commit_count, 0);
wmsum_init(&zs->zil_commit_writer_count, 0); wmsum_init(&zs->zil_commit_writer_count, 0);
wmsum_init(&zs->zil_commit_error_count, 0);
wmsum_init(&zs->zil_commit_stall_count, 0);
wmsum_init(&zs->zil_commit_suspend_count, 0);
wmsum_init(&zs->zil_itx_count, 0); wmsum_init(&zs->zil_itx_count, 0);
wmsum_init(&zs->zil_itx_indirect_count, 0); wmsum_init(&zs->zil_itx_indirect_count, 0);
wmsum_init(&zs->zil_itx_indirect_bytes, 0); wmsum_init(&zs->zil_itx_indirect_bytes, 0);
@ -382,6 +388,9 @@ zil_sums_fini(zil_sums_t *zs)
{ {
wmsum_fini(&zs->zil_commit_count); wmsum_fini(&zs->zil_commit_count);
wmsum_fini(&zs->zil_commit_writer_count); wmsum_fini(&zs->zil_commit_writer_count);
wmsum_fini(&zs->zil_commit_error_count);
wmsum_fini(&zs->zil_commit_stall_count);
wmsum_fini(&zs->zil_commit_suspend_count);
wmsum_fini(&zs->zil_itx_count); wmsum_fini(&zs->zil_itx_count);
wmsum_fini(&zs->zil_itx_indirect_count); wmsum_fini(&zs->zil_itx_indirect_count);
wmsum_fini(&zs->zil_itx_indirect_bytes); wmsum_fini(&zs->zil_itx_indirect_bytes);
@ -406,6 +415,12 @@ zil_kstat_values_update(zil_kstat_values_t *zs, zil_sums_t *zil_sums)
wmsum_value(&zil_sums->zil_commit_count); wmsum_value(&zil_sums->zil_commit_count);
zs->zil_commit_writer_count.value.ui64 = zs->zil_commit_writer_count.value.ui64 =
wmsum_value(&zil_sums->zil_commit_writer_count); wmsum_value(&zil_sums->zil_commit_writer_count);
zs->zil_commit_error_count.value.ui64 =
wmsum_value(&zil_sums->zil_commit_error_count);
zs->zil_commit_stall_count.value.ui64 =
wmsum_value(&zil_sums->zil_commit_stall_count);
zs->zil_commit_suspend_count.value.ui64 =
wmsum_value(&zil_sums->zil_commit_suspend_count);
zs->zil_itx_count.value.ui64 = zs->zil_itx_count.value.ui64 =
wmsum_value(&zil_sums->zil_itx_count); wmsum_value(&zil_sums->zil_itx_count);
zs->zil_itx_indirect_count.value.ui64 = zs->zil_itx_indirect_count.value.ui64 =
@ -2823,6 +2838,7 @@ zil_commit_writer_stall(zilog_t *zilog)
* (which is achieved via the txg_wait_synced() call). * (which is achieved via the txg_wait_synced() call).
*/ */
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
ZIL_STAT_BUMP(zilog, zil_commit_stall_count);
txg_wait_synced(zilog->zl_dmu_pool, 0); txg_wait_synced(zilog->zl_dmu_pool, 0);
ASSERT(list_is_empty(&zilog->zl_lwb_list)); ASSERT(list_is_empty(&zilog->zl_lwb_list));
} }
@ -3592,6 +3608,7 @@ zil_commit(zilog_t *zilog, uint64_t foid)
* semantics, and avoid calling those functions altogether. * semantics, and avoid calling those functions altogether.
*/ */
if (zilog->zl_suspend > 0) { if (zilog->zl_suspend > 0) {
ZIL_STAT_BUMP(zilog, zil_commit_suspend_count);
txg_wait_synced(zilog->zl_dmu_pool, 0); txg_wait_synced(zilog->zl_dmu_pool, 0);
return; return;
} }
@ -3645,10 +3662,12 @@ zil_commit_impl(zilog_t *zilog, uint64_t foid)
* implications, but the expectation is for this to be * implications, but the expectation is for this to be
* an exceptional case, and shouldn't occur often. * an exceptional case, and shouldn't occur often.
*/ */
ZIL_STAT_BUMP(zilog, zil_commit_error_count);
DTRACE_PROBE2(zil__commit__io__error, DTRACE_PROBE2(zil__commit__io__error,
zilog_t *, zilog, zil_commit_waiter_t *, zcw); zilog_t *, zilog, zil_commit_waiter_t *, zcw);
txg_wait_synced(zilog->zl_dmu_pool, 0); txg_wait_synced(zilog->zl_dmu_pool, 0);
} else if (wtxg != 0) { } else if (wtxg != 0) {
ZIL_STAT_BUMP(zilog, zil_commit_suspend_count);
txg_wait_synced(zilog->zl_dmu_pool, wtxg); txg_wait_synced(zilog->zl_dmu_pool, wtxg);
} }