block: add logging facility for long standing IO requests
cover block: add logging facility for long standing IO requests
There are severe delays with IO requests processing if QEMU is running in virtual machine or over software defined storage. Such delays potentially results in unpredictable guest behavior. For example, guests over IDE or SATA drive could remount filesystem read-only if write is performed longer than 10 seconds.
Such reports are very complex to process. Some good starting point for this seems quite reasonable. This patch provides one. It adds logging of such potentially dangerous long IO operations.
PATCH 1/2 block/block-backend: add converter from BlockAcctStats to BlockBackend
Right now BlockAcctStats is always reside on BlockBackend. This structure is not used in any other place. Thus we are able to create a converter from one pointer to another.
--- block/block-backend.c | 5 +++++ include/sysemu/block-backend.h | 1 + 2 files changed, 6 insertions(+)
block/block-backend.c +5 -0
@@ -2143,6 +2143,11 @@ BlockAcctStats *blk_get_stats(BlockBackend *blk)return &blk->stats; } +BlockBackend *blk_stats2blk (BlockAcctStats *s) +{ +return container_of (s, BlockBackend, stats); +} +void *blk_aio_get (const AIOCBInfo *aiocb_info,BlockBackend *blk,BlockCompletionFunc *cb,void *opaque) {
include/sysemu/block-backend.h +1 -0
@@ -227,6 +227,7 @@ void blk_add_insert_bs_notifier(BlockBackend *blk, Notifier *notify);void blk_io_plug (BlockBackend *blk);void blk_io_unplug (BlockBackend *blk);BlockAcctStats *blk_get_stats (BlockBackend *blk); +BlockBackend *blk_stats2blk (BlockAcctStats *stats);BlockBackendRootState *blk_get_root_state (BlockBackend *blk);void blk_update_root_state (BlockBackend *blk);bool blk_get_detect_zeroes_from_root_state (BlockBackend *blk);
Signature · 1 line
2.17.1
Re: [PATCH 1/2] block/block-backend: add converter from BlockAcctStats to BlockBackend
Quoted 35 lines · Denis V. Lunev
Right now BlockAcctStats is always reside on BlockBackend. This structure is not used in any other place. Thus we are able to create a converter from one pointer to another.
Signed-off-by Denis V. Lunev <[email protected]>CC Vladimir Sementsov-Ogievskiy <[email protected]>CC Kevin Wolf <[email protected]>CC Max Reitz <[email protected]>--- block/block-backend.c | 5 +++++ include/sysemu/block-backend.h | 1 + 2 files changed, 6 insertions(+)block/block-backend.c +5
-0@@ -2143,6 +2143,11 @@ BlockAcctStats *blk_get_stats(BlockBackend *blk)return &blk->stats; } +BlockBackend *blk_stats2blk (BlockAcctStats *s) +{ +return container_of (s, BlockBackend, stats); +} +void *blk_aio_get (const AIOCBInfo *aiocb_info,BlockBackend *blk,BlockCompletionFunc *cb,void *opaque) {include/sysemu/block-backend.h +0
-0
include/sysemu/block-backend.h void blk_add_insert_bs_notifier(BlockBackend *blk, Notifier *notify); · line 233
Quoted 7 lines
include/sysemu/block-backend.h +1
-0@@ -227,6 +227,7 @@ void blk_add_insert_bs_notifier(BlockBackend *blk, Notifier *notify);void blk_io_plug (BlockBackend *blk);void blk_io_unplug (BlockBackend *blk);BlockAcctStats *blk_get_stats (BlockBackend *blk); +BlockBackend *blk_stats2blk (BlockAcctStats *stats);BlockBackendRootState *blk_get_root_state (BlockBackend *blk);void blk_update_root_state (BlockBackend *blk);bool blk_get_detect_zeroes_from_root_state (BlockBackend *blk);
Signature · 2 lines
Best regards, Vladimir
PATCH 2/2 block: add logging facility for long standing IO requests
In reply to Denis V. LunevThere are severe delays with IO requests processing if QEMU is running in virtual machine or over software defined storage. Such delays potentially results in unpredictable guest behavior. For example, guests over IDE or SATA drive could remount filesystem read-only if write is performed longer than 10 seconds.
Such reports are very complex to process. Some good starting point for this seems quite reasonable. This patch provides one. It adds logging of such potentially dangerous long IO operations.
--- block/accounting.c | 59 +++++++++++++++++++++++++++++++++++++- blockdev.c | 7 ++++- include/block/accounting.h | 5 +++- 3 files changed, 68 insertions(+), 3 deletions(-)
block/accounting.c +58 -1
@@ -27,7 +27,9 @@#include "block/accounting.h" #include "block/block_int.h" #include "qemu/timer.h" +#include "qemu/log.h" #include "sysemu/qtest.h" +#include "sysemu/block-backend.h" static QEMUClockType clock_type =QEMU_CLOCK_REALTIME ;static const int qtest_latency_ns =NANOSECONDS_PER_SECOND /1000 ;
@@ -41,10 +43,11 @@ void block_acct_init(BlockAcctStats *stats) }void block_acct_setup (BlockAcctStats *stats,bool account_invalid, -bool account_failed) +bool account_failed,unsigned latency_log_threshold_ms) { stats->account_invalid = account_invalid; stats->account_failed = account_failed; + stats->latency_log_threshold_ms = latency_log_threshold_ms; }void block_acct_cleanup (BlockAcctStats *stats)
@@ -182,6 +185,58 @@ void block_latency_histograms_clear(BlockAcctStats *stats) } } +static const char *block_account_type (enum BlockAcctType type) +{ +switch (type) { +case BLOCK_ACCT_READ : +return "READ" ; +case BLOCK_ACCT_WRITE : +return "WRITE" ; +case BLOCK_ACCT_FLUSH : +return "DISCARD" ; +case BLOCK_ACCT_UNMAP : +return "TRUNCATE" ; +case BLOCK_ACCT_NONE : +return "NONE" ; +case BLOCK_MAX_IOTYPE : +break ; + } +return "UNKNOWN" ; +} + +static void block_acct_report_long (BlockAcctStats *stats, +BlockAcctCookie *cookie,int64_t latency_ns) +{ +unsigned latency_ms = latency_ns /1000000 ; +int64_t start_time_host_s; +char buf[64 ]; +struct tm t; +BlockDriverState *bs; + +if (cookie->type ==BLOCK_ACCT_NONE ) { +return ; + } +if (stats->latency_log_threshold_ms ==0 ) { +return ; + } +if (latency_ms < stats->latency_log_threshold_ms) { +return ; + } + + start_time_host_s = cookie->start_time_ns /1000000000ull ; +strftime (buf,sizeof (buf),"%m-%d %H:%M:%S" , +localtime_r (&start_time_host_s, &t)); + + bs =blk_bs (blk_stats2blk (stats)); +qemu_log ("long %s[%ld] IO request: %d.03%d since %s.%03d bs: %s(%s, %s)\n" , +block_account_type (cookie->type), cookie->bytes, + (int )(latency_ms /1000 ), (int )(latency_ms %1000 ), buf, + (int )((cookie->start_time_ns /1000000 ) %1000 ), + bs ==NULL ?"unknown" :bdrv_get_node_name (bs), + bs ==NULL ?"unknown" :bdrv_get_format_name (bs), + bs ==NULL ?"unknown" : bs->filename); +} +static void block_account_one_io (BlockAcctStats *stats,BlockAcctCookie *cookie,bool failed) {
@@ -222,6 +277,8 @@ static void block_account_one_io(BlockAcctStats *stats, BlockAcctCookie *cookie,qemu_mutex_unlock (&stats->lock); +block_acct_report_long (stats, cookie, latency_ns); + cookie->type =BLOCK_ACCT_NONE ; }
blockdev.c +6 -1
@@ -625,7 +625,8 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts, bs->detect_zeroes = detect_zeroes; -block_acct_setup (blk_get_stats (blk), account_invalid, account_failed); +block_acct_setup (blk_get_stats (blk), account_invalid, account_failed, +qemu_opt_get_number (opts,"latency-log-threshold" ,0 ));if (!parse_stats_intervals (blk_get_stats (blk), interval_list, errp)) {blk_unref (blk);
@@ -3740,6 +3741,10 @@ QemuOptsList qemu_common_drive_opts = { .type = QEMU_OPT_BOOL, .help = "whether to account for failed I/O operations " "in the statistics", + },{ + .name = "latency-log-threshold", + .type = QEMU_OPT_STRING, + .help = "threshold for long I/O report (disabled if <=0), in ms", }, {/* end of list */ } },
include/block/accounting.h +4 -1
@@ -91,6 +91,9 @@ struct BlockAcctStats {bool account_invalid;bool account_failed;BlockLatencyHistogram latency_histogram[BLOCK_MAX_IOTYPE ]; + +/* Threshold for long I/O detection, ms */ +unsigned latency_log_threshold_ms; }; typedef struct BlockAcctCookie {
@@ -101,7 +104,7 @@ typedef struct BlockAcctCookie {void block_acct_init (BlockAcctStats *stats);void block_acct_setup (BlockAcctStats *stats,bool account_invalid, -bool account_failed); +bool account_failed,unsigned latency_log_threshold_ms);void block_acct_cleanup (BlockAcctStats *stats);void block_acct_add_interval (BlockAcctStats *stats,unsigned interval_length); BlockAcctTimedStats *block_acct_interval_next(BlockAcctStats *stats,
Signature · 1 line
2.17.1
Quoted 85 lines · Denis V. Lunev
There are severe delays with IO requests processing if QEMU is running in virtual machine or over software defined storage. Such delays potentially results in unpredictable guest behavior. For example, guests over IDE or SATA drive could remount filesystem read-only if write is performed longer than 10 seconds.
Such reports are very complex to process. Some good starting point for this seems quite reasonable. This patch provides one. It adds logging of such potentially dangerous long IO operations.
Signed-off-by Denis V. Lunev <[email protected]>CC Vladimir Sementsov-Ogievskiy <[email protected]>CC Kevin Wolf <[email protected]>CC Max Reitz <[email protected]>--- block/accounting.c | 59 +++++++++++++++++++++++++++++++++++++- blockdev.c | 7 ++++- include/block/accounting.h | 5 +++- 3 files changed, 68 insertions(+), 3 deletions(-)block/accounting.c +38
-1@@ -27,7 +27,9 @@#include "block/accounting.h" #include "block/block_int.h" #include "qemu/timer.h" +#include "qemu/log.h" #include "sysemu/qtest.h" +#include "sysemu/block-backend.h" static QEMUClockType clock_type =QEMU_CLOCK_REALTIME ;static const int qtest_latency_ns =NANOSECONDS_PER_SECOND /1000 ;@@ -41,10 +43,11 @@ void block_acct_init(BlockAcctStats *stats) }void block_acct_setup (BlockAcctStats *stats,bool account_invalid, -bool account_failed) +bool account_failed,unsigned latency_log_threshold_ms) { stats->account_invalid = account_invalid; stats->account_failed = account_failed; + stats->latency_log_threshold_ms = latency_log_threshold_ms; }void block_acct_cleanup (BlockAcctStats *stats)@@ -182,6 +185,58 @@ void block_latency_histograms_clear(BlockAcctStats *stats) } } +static const char *block_account_type (enum BlockAcctType type) +{ +switch (type) { +case BLOCK_ACCT_READ : +return "READ" ; +case BLOCK_ACCT_WRITE : +return "WRITE" ; +case BLOCK_ACCT_FLUSH : +return "DISCARD" ; +case BLOCK_ACCT_UNMAP : +return "TRUNCATE" ; +case BLOCK_ACCT_NONE : +return "NONE" ; +case BLOCK_MAX_IOTYPE : +break ; + } +return "UNKNOWN" ; +} + +static void block_acct_report_long (BlockAcctStats *stats, +BlockAcctCookie *cookie,int64_t latency_ns) +{ +unsigned latency_ms = latency_ns /1000000 ; +int64_t start_time_host_s; +char buf[64 ]; +struct tm t; +BlockDriverState *bs; + +if (cookie->type ==BLOCK_ACCT_NONE ) { +return ; + } +if (stats->latency_log_threshold_ms ==0 ) { +return ; + }
block/accounting.c void block_latency_histograms_clear(BlockAcctStats *stats) · line 231
Quoted 10 lines
block/accounting.c +10
-0+if (latency_ms < stats->latency_log_threshold_ms) { +return ; + } + + start_time_host_s = cookie->start_time_ns /1000000000ull ; +strftime (buf,sizeof (buf),"%m-%d %H:%M:%S" , +localtime_r (&start_time_host_s, &t)); + + bs =blk_bs (blk_stats2blk (stats)); + qemu_log("long %s[%ld] IO request: %d.03%d since %s.%03d bs: %s(%s, %s)\n",
Better use PRId64 for cookie->bytes
s/.03%d/.%03d/
with this fixed:
block/accounting.c void block_latency_histograms_clear(BlockAcctStats *stats) · line 237
Quoted 6 lines
block/accounting.c +6
-0+block_account_type (cookie->type), cookie->bytes, + (int )(latency_ms /1000 ), (int )(latency_ms %1000 ), buf, + (int )((cookie->start_time_ns /1000000 ) %1000 ), + bs ==NULL ?"unknown" :bdrv_get_node_name (bs), + bs ==NULL ?"unknown" :bdrv_get_format_name (bs), + bs == NULL ? "unknown" : bs->filename);
may add blk_name as well
Quoted 19 lines
block/accounting.c +4
-0+} +static void block_account_one_io (BlockAcctStats *stats,BlockAcctCookie *cookie,bool failed) {@@ -222,6 +277,8 @@ static void block_account_one_io(BlockAcctStats *stats, BlockAcctCookie *cookie,qemu_mutex_unlock (&stats->lock); +block_acct_report_long (stats, cookie, latency_ns); + cookie->type =BLOCK_ACCT_NONE ; }blockdev.c +0
-0
blockdev.c static BlockBackend *blockdev_init(const char *file, QDict *bs_opts, · line 629
Quoted 6 lines
blockdev.c +2
-1@@ -625,7 +625,8 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts, bs->detect_zeroes = detect_zeroes; -block_acct_setup (blk_get_stats (blk), account_invalid, account_failed); +block_acct_setup (blk_get_stats (blk), account_invalid, account_failed, +qemu_opt_get_number (opts,"latency-log-threshold" ,0 ));
Could we use non-zero default, so the option to be enabled by default?
blockdev.c QemuOptsList qemu_common_drive_opts = { · line 3747
Quoted 11 lines
blockdev.c +4
-0if (!parse_stats_intervals (blk_get_stats (blk), interval_list, errp)) {blk_unref (blk);@@ -3740,6 +3741,10 @@ QemuOptsList qemu_common_drive_opts = { .type = QEMU_OPT_BOOL, .help = "whether to account for failed I/O operations " "in the statistics", + },{ + .name = "latency-log-threshold", + .type = QEMU_OPT_STRING, + .help = "threshold for long I/O report (disabled if <=0), in ms",
I'd say if ==0. Anyway it's unsigned..
Quoted 27 lines
blockdev.c +0
-0}, {/* end of list */ } },include/block/accounting.h +4
-1@@ -91,6 +91,9 @@ struct BlockAcctStats {bool account_invalid;bool account_failed;BlockLatencyHistogram latency_histogram[BLOCK_MAX_IOTYPE ]; + +/* Threshold for long I/O detection, ms */ +unsigned latency_log_threshold_ms; }; typedef struct BlockAcctCookie {@@ -101,7 +104,7 @@ typedef struct BlockAcctCookie {void block_acct_init (BlockAcctStats *stats);void block_acct_setup (BlockAcctStats *stats,bool account_invalid, -bool account_failed); +bool account_failed,unsigned latency_log_threshold_ms);void block_acct_cleanup (BlockAcctStats *stats);void block_acct_add_interval (BlockAcctStats *stats,unsigned interval_length); BlockAcctTimedStats *block_acct_interval_next(BlockAcctStats *stats,
Signature · 2 lines
Best regards, Vladimir
block/accounting.c void block_latency_histograms_clear(BlockAcctStats *stats) · line 210
Quoted 4 lines · Denis V. Lunev
block/accounting.c +4
-0+static void block_acct_report_long (BlockAcctStats *stats, +BlockAcctCookie *cookie,int64_t latency_ns) +{ +unsigned latency_ms = latency_ns /1000000 ;
Please use SCALE_MS.
Quoted 6 lines
block/accounting.c +6
-0+int64_t start_time_host_s; +char buf[64 ]; +struct tm t; +BlockDriverState *bs; + +if (cookie->type ==BLOCK_ACCT_NONE ) {
block/accounting.c void block_latency_histograms_clear(BlockAcctStats *stats) · line 226
Quoted 10 lines
block/accounting.c +10
-0+return ; + } +if (stats->latency_log_threshold_ms ==0 ) { +return ; + } +if (latency_ms < stats->latency_log_threshold_ms) { +return ; + } + + start_time_host_s = cookie->start_time_ns /1000000000ull ;
Please use NANOSECONDS_PER_SECOND.
Quoted 15 lines
blockdev.c +2
-1@@ -625,7 +625,8 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts, bs->detect_zeroes = detect_zeroes; -block_acct_setup (blk_get_stats (blk), account_invalid, account_failed); +block_acct_setup (blk_get_stats (blk), account_invalid, account_failed, +qemu_opt_get_number (opts,"latency-log-threshold" ,0 ));if (!parse_stats_intervals (blk_get_stats (blk), interval_list, errp)) {blk_unref (blk);
blockdev.c QemuOptsList qemu_common_drive_opts = { · line 3746
Quoted 6 lines
blockdev.c +3
-0@@ -3740,6 +3741,10 @@ QemuOptsList qemu_common_drive_opts = { .type = QEMU_OPT_BOOL, .help = "whether to account for failed I/O operations " "in the statistics", + },{ + .name = "latency-log-threshold", + .type = QEMU_OPT_STRING,
QEMU_OPT_NUMBER?
Quoted 6 lines · Stefan Hajnoczi
block/accounting.c void block_latency_histograms_clear(BlockAcctStats *stats) · line 210
Quoted 4 lines · Denis V. Lunev
block/accounting.c +4
-0+static void block_acct_report_long (BlockAcctStats *stats, +BlockAcctCookie *cookie,int64_t latency_ns) +{ +unsigned latency_ms = latency_ns /1000000 ;Please use
SCALE_MS.
sure
Quoted 10 lines
Quoted 10 lines
block/accounting.c +9
-0+int64_t start_time_host_s; +char buf[64 ]; +struct tm t; +BlockDriverState *bs; + +if (cookie->type ==BLOCK_ACCT_NONE ) { +return ; + } +if (stats->latency_log_threshold_ms ==0 ) {
Quoted 8 lines
block/accounting.c void block_latency_histograms_clear(BlockAcctStats *stats) · line 226
Quoted 7 lines
block/accounting.c +7
-0+return ; + } +if (latency_ms < stats->latency_log_threshold_ms) { +return ; + } + + start_time_host_s = cookie->start_time_ns /1000000000ull ;Please use
NANOSECONDS_PER_SECOND.
sure
Quoted 12 lines
Quoted 12 lines
blockdev.c +2
-1@@ -625,7 +625,8 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts, bs->detect_zeroes = detect_zeroes; -block_acct_setup (blk_get_stats (blk), account_invalid, account_failed); +block_acct_setup (blk_get_stats (blk), account_invalid, account_failed, +qemu_opt_get_number (opts,"latency-log-threshold" ,0 ));
Quoted 10 lines
blockdev.c QemuOptsList qemu_common_drive_opts = { · line 3746
Quoted 9 lines
blockdev.c +3
-0if (!parse_stats_intervals (blk_get_stats (blk), interval_list, errp)) {blk_unref (blk);@@ -3740,6 +3741,10 @@ QemuOptsList qemu_common_drive_opts = { .type = QEMU_OPT_BOOL, .help = "whether to account for failed I/O operations " "in the statistics", + },{ + .name = "latency-log-threshold", + .type = QEMU_OPT_STRING,
QEMU_OPT_NUMBER?
good catch! we have this parameter on by default (10 seconds) and never passed it from the command line :)
Thus should we keep this off be default or on? :)
Den
Quoted 10 lines · Denis V. Lunev
Quoted 9 lines · Stefan Hajnoczi
blockdev.c QemuOptsList qemu_common_drive_opts = { · line 3746
Quoted 7 lines · Denis V. Lunev
blockdev.c +3
-0@@ -3740,6 +3741,10 @@ QemuOptsList qemu_common_drive_opts = { .type = QEMU_OPT_BOOL, .help = "whether to account for failed I/O operations " "in the statistics", + },{ + .name = "latency-log-threshold", + .type = QEMU_OPT_STRING,
QEMU_OPT_NUMBER?
Quoted 4 lines
good catch! we have this parameter on by default (10 seconds) and never passed it from the command line :)
Thus should we keep this off be default or on? :)
On by default sounds good. It adds useful information.
Stefan
Quoted 6 lines · Denis V. Lunev
There are severe delays with IO requests processing if QEMU is running in virtual machine or over software defined storage. Such delays potentially results in unpredictable guest behavior. For example, guests over IDE or SATA drive could remount filesystem read-only if write is performed longer than 10 seconds.
Quoted 8 lines
Such reports are very complex to process. Some good starting point for this seems quite reasonable. This patch provides one. It adds logging of such potentially dangerous long IO operations.
Signed-off-by Denis V. Lunev <[email protected]>CC Vladimir Sementsov-Ogievskiy <[email protected]>CC Kevin Wolf <[email protected]>CC Max Reitz <[email protected]>
This looks useful. It is indeed hard to diagnose soft lockups, I/O timeouts, etc inside the guest :). QEMU should print more info. Thanks for doing this!
Stefan
Quoted 6 lines · Denis V. Lunev
There are severe delays with IO requests processing if QEMU is running in virtual machine or over software defined storage. Such delays potentially results in unpredictable guest behavior. For example, guests over IDE or SATA drive could remount filesystem read-only if write is performed longer than 10 seconds.
Quoted 8 lines
Such reports are very complex to process. Some good starting point for this seems quite reasonable. This patch provides one. It adds logging of such potentially dangerous long IO operations.
Signed-off-by Denis V. Lunev <[email protected]>CC Vladimir Sementsov-Ogievskiy <[email protected]>CC Kevin Wolf <[email protected]>CC Max Reitz <[email protected]>
If I understand correctly this only reports completed I/Os, so if the host hasn't given up on an I/O request yet then QEMU will not report it is taking a long time. In the meantime the guest could start printing timeout errors.
I think this patch series is good as it is. In the future maybe a QMP command that lists in-flight I/O requests would be nice. That helps when troubleshooting I/Os that are hung.
Stefan
Quoted 15 lines · Stefan Hajnoczi
Quoted 6 lines · Denis V. Lunev
There are severe delays with IO requests processing if QEMU is running in virtual machine or over software defined storage. Such delays potentially results in unpredictable guest behavior. For example, guests over IDE or SATA drive could remount filesystem read-only if write is performed longer than 10 seconds.
Quoted 8 lines
Such reports are very complex to process. Some good starting point for this seems quite reasonable. This patch provides one. It adds logging of such potentially dangerous long IO operations.
Signed-off-by Denis V. Lunev <[email protected]>CC Vladimir Sementsov-Ogievskiy <[email protected]>CC Kevin Wolf <[email protected]>CC Max Reitz <[email protected]>
Quoted 8 lines
If I understand correctly this only reports completed I/Os, so if the host hasn't given up on an I/O request yet then QEMU will not report it is taking a long time. In the meantime the guest could start printing timeout errors.
I think this patch series is good as it is. In the future maybe a QMP command that lists in-flight I/O requests would be nice. That helps when troubleshooting I/Os that are hung.
We could dump requests at block level, we do have proper lists, but the would be a little bit different.
Anyway, I __DO__ like this idea, I have missed this somehow :)
Thank you for the suggestion,
Den