[Devel] [PATCH RHEL7 COMMIT] fs/fuse kio: implement fstat_lat statistics info
Konstantin Khorenko
khorenko at virtuozzo.com
Tue May 21 18:47:46 MSK 2019
The commit is pushed to "branch-rh7-3.10.0-957.12.2.vz7.96.x-ovz" and will appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-957.12.2.vz7.96.1
------>
commit 985998e70e6a3e1d9507de69035f39f8752d9459
Author: Pavel Butsykin <pbutsykin at virtuozzo.com>
Date: Tue May 21 18:47:44 2019 +0300
fs/fuse kio: implement fstat_lat statistics info
This statistic shows latency of requests for each open kio file. Latency is
displayed for 99 and 95 percentile in the following time ranges: 20ms, 50ms,
100ms, 200ms, 500ms. In turn, the request latency is detailed in three parts:
pending_lat - time spent in the queue
net_lat - time spent in the network
io_lat - disk operation time
example:
# rw open/age inactive net_lat net_lat_global io_lat io_lat_global plat plat_global path
rw 2/6360 0 (5% > 500ms, 1% > 500ms) (5% > 500ms, 1% > 500ms) (95% <= 20ms, 99% <= 50ms) (95% <= 20ms, 99% <= 50ms) (95% <= 20ms, 99% <= 20ms) (95% <= 20ms, 99% <= 20ms) /fio_test/randasyncread_jobs24_4k.0.0
rw 2/6360 0 (5% > 500ms, 1% > 500ms) (5% > 500ms, 1% > 500ms) (95% <= 20ms, 99% <= 50ms) (95% <= 20ms, 99% <= 20ms) (95% <= 20ms, 99% <= 20ms) (95% <= 20ms, 99% <= 20ms) /fio_test/randasyncread_jobs24_4k.1.0
Signed-off-by: Pavel Butsykin <pbutsykin at virtuozzo.com>
=====================
Patchset description:
Statistics for vstorage kernel fast-path
Vstorage provides plenty of statistics information via 'vstorage -c cl mnt-top',
but when it comes kernel fast-path, it doesn't work. All because mnt-top
command collects information from .vstorage.info directory, where vstorage-mount
provides a lot of different statistical information in the form of files, but
it was not implemented for for fast-path.
This patch-set is aimed to implementation of the support of some statistics
inforamtion files from .vstorage.info:
cs_stats
fstat
fstat_lat
iostat
requests
Which will be located at "/sys/fs/fuse/connections/*mnt_id*/kio_stat/". This
will be enough to maintain vstorage mnt-top command for fast-path mount points.
https://pmc.acronis.com/browse/VSTOR-20979
Pavel Butsykin (15):
fs/fuse: add conn_ctl to fuse_conn struct
fs/fuse kio: create sysfs stat directory
fs/fuse kio: implement iostat
fs/fuse kio: make common interface pcs_kio_file_list() for listing kio files
fs/fuse kio: make common interface pcs_kio_req_list() for listing kio reqs
fs/fuse kio: add retry counter for kio requests
fs/fuse kio: implement pcs_strerror()
fs/fuse kio: implement requests statistics
fs/fuse kio: implement fstat statistics info
fs/fuse kio: implement fstat_lat statistics info
fs/fuse kio: remove useless pcs_cs initialization
fs/fuse kio: implement cs statistics accounting
fs/fuse kio: convert rpc state id to string
fs/fuse kio: implement cs_stats statistics info
fs/fuse kio: add locked cs_get_avg_in_flight()
---
fs/fuse/kio/pcs/fuse_stat.c | 183 +++++++++++++++++++++++++++++++++++++++++++-
fs/fuse/kio/pcs/fuse_stat.h | 30 +++++++-
fs/fuse/kio/pcs/pcs_map.c | 2 +-
3 files changed, 212 insertions(+), 3 deletions(-)
diff --git a/fs/fuse/kio/pcs/fuse_stat.c b/fs/fuse/kio/pcs/fuse_stat.c
index 85c8f7e04413..452bb4e9664c 100644
--- a/fs/fuse/kio/pcs/fuse_stat.c
+++ b/fs/fuse/kio/pcs/fuse_stat.c
@@ -7,6 +7,44 @@
extern struct super_block *fuse_control_sb;
+static u64 lat_oreder_list[] = {
+ [LAT_ORDER1] = 20 * USEC_PER_MSEC,
+ [LAT_ORDER2] = 50 * USEC_PER_MSEC,
+ [LAT_ORDER3] = 100 * USEC_PER_MSEC,
+ [LAT_ORDER4] = 200 * USEC_PER_MSEC,
+ [LAT_ORDER5] = 500 * USEC_PER_MSEC,
+};
+
+
+static void fuse_latency_update(struct fuse_lat_stat *s, u64 lat)
+{
+ int i;
+
+ s->count++;
+
+ for (i = LAT_ORDER1; i <= LAT_ORDER5; i++) {
+ if (likely(lat <= lat_oreder_list[i])) {
+ s->lat[i]++;
+ return;
+ }
+ }
+ s->lat[LAT_ORDER_OTHER]++;
+ return;
+}
+
+static inline void fuse_latency_count(struct fuse_lat_cnt *c, u64 val)
+{
+ fuse_latency_update(&c->curr, val);
+ fuse_latency_update(&c->glob, val);
+}
+
+static inline void fuse_latency_cnt_up(struct fuse_lat_cnt *c)
+{
+ BUILD_BUG_ON(sizeof(c->last) != sizeof(c->curr));
+
+ memcpy(&c->last, &c->curr, sizeof(c->last));
+ memset(&c->curr, 0, sizeof(c->curr));
+}
static const char *fuse_kio_op_name(unsigned opcode)
{
@@ -96,12 +134,142 @@ static inline unsigned long long fuse_val_cnt_max(struct fuse_val_cnt const* c)
#define CNT_MAX(c) fuse_val_cnt_max(&(c))
+#define MAX_PERCENT 100
+static int latency_npercl_format(struct fuse_lat_stat *s, u8 percl, char *buf,
+ size_t sz)
+{
+ u64 cnt = s->count, nper;
+ int i;
+
+ BUG_ON(percl > MAX_PERCENT);
+
+ if (!cnt)
+ return 0;
+
+ nper = cnt * (MAX_PERCENT - percl);
+ if (s->lat[LAT_ORDER_OTHER] * MAX_PERCENT >= nper)
+ return scnprintf(buf, sz, "%u%% > %llums", MAX_PERCENT - percl,
+ lat_oreder_list[LAT_ORDER_OTHER - 1] / 1000);
+
+ for(i = LAT_ORDER_OTHER - 1; i >= 0; i--) {
+ if (s->lat[i] * MAX_PERCENT >= nper)
+ return scnprintf(buf, sz, "%u%% <= %llums", percl,
+ lat_oreder_list[i] / 1000);
+ }
+ BUG();
+ return 0;
+}
+
+#define LAT_LINE_MAX 27
+static void latency_percl_print(struct fuse_lat_stat *s, struct seq_file *m)
+{
+ int n, icnt = LAT_LINE_MAX;
+ char buf[LAT_LINE_MAX];
+
+ seq_printf(m, "(");
+ n = latency_npercl_format(s, 95, buf, sizeof(buf) - 1);
+ if (n) {
+ seq_printf(m,"%.*s, ", n, buf);
+ icnt -= n;
+ icnt -= 2;
+ }
+ n = latency_npercl_format(s, 99, buf, sizeof(buf) - 1);
+ if (n) {
+ seq_printf(m, "%.*s", n, buf);
+ icnt -= n;
+ }
+
+ WARN_ON(icnt < 0);
+ seq_printf(m, ")%*s", max(icnt, 0), "");
+}
+
+static void fuse_kio_fstat_lat_itr(struct fuse_file *ff,
+ struct pcs_dentry_info *di, void *ctx)
+{
+ struct seq_file *m = ctx;
+ struct fuse_io_cnt *fstat = &di->stat;
+ umode_t mode = di->inode->inode.i_mode;
+ abs_time_t now = jiffies;
+ struct fuse_lat_cnt io_lat;
+ struct fuse_lat_cnt net_lat;
+ struct fuse_lat_cnt pending_lat;
+
+ seq_printf(m, "%s%s %7u/%-7llu %-5u",
+ mode & S_IRUGO ? "r": "", mode & S_IWUGO ? "w": "",
+ atomic_read(&ff->count), (now - di->stat.created_ts) / 1000, 0);
+
+ spin_lock(&fstat->lock);
+ io_lat = fstat->io_lat;
+ net_lat = fstat->net_lat;
+ pending_lat = fstat->pending_lat;
+ spin_unlock(&fstat->lock);
+
+ latency_percl_print(&net_lat.last, m);
+ latency_percl_print(&net_lat.glob, m);
+
+ latency_percl_print(&io_lat.last, m);
+ latency_percl_print(&io_lat.glob, m);
+
+ latency_percl_print(&pending_lat.last, m);
+ latency_percl_print(&pending_lat.glob, m);
+
+ seq_dentry(m, ff->ff_dentry, "");
+ seq_putc(m, '\n');
+}
+
+static int pcs_fuse_fstat_lat_show(struct seq_file *m, void *v)
+{
+ struct inode *inode = m->private;
+ struct pcs_fuse_stat *stat;
+ struct fuse_conn *fc;
+
+ if (!inode)
+ return 0;
+
+ mutex_lock(&fuse_mutex);
+ stat = inode->i_private;
+ if (!stat) {
+ mutex_unlock(&fuse_mutex);
+ return 0;
+ }
+
+ seq_printf(m, "# rw open/age inactive net_lat\t\t\t net_lat_global\t\t io_lat\t\t io_lat_global\t\t plat\t\t\t plat_global\t\t path\n");
+
+ fc = container_of(stat, struct pcs_fuse_cluster, cc.stat)->fc;
+ if (fc) {
+ spin_lock(&fc->lock);
+ pcs_kio_file_list(fc, fuse_kio_fstat_lat_itr, m);
+ spin_unlock(&fc->lock);
+ }
+ mutex_unlock(&fuse_mutex);
+
+
+ return 0;
+}
+
+static int pcs_fuse_fstat_lat_open(struct inode *inode, struct file *file)
+{
+ return single_open(file, pcs_fuse_fstat_lat_show, inode);
+}
+
+static const struct file_operations pcs_fuse_fstat_lat_ops = {
+ .owner = THIS_MODULE,
+ .open = pcs_fuse_fstat_lat_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+};
+
static void fuse_fstat_up_itr(struct fuse_file *ff, struct pcs_dentry_info *di,
void *ctx)
{
struct fuse_io_cnt *fstat = &di->stat;
spin_lock(&fstat->lock);
+ fuse_latency_cnt_up(&fstat->io_lat);
+ fuse_latency_cnt_up(&fstat->net_lat);
+ fuse_latency_cnt_up(&fstat->pending_lat);
+
fuse_val_cnt_up(&fstat->io.read_bytes);
fuse_val_cnt_up(&fstat->io.write_bytes);
fuse_val_cnt_up(&fstat->io.flush_cnt);
@@ -316,7 +484,8 @@ struct fuse_val_cnt *req_stat_entry(struct pcs_fuse_io_stat *io, u32 type)
return NULL;
}
-void pcs_fuse_stat_io_count(struct pcs_int_request *ireq, struct pcs_msg *resp)
+void pcs_fuse_stat_io_count(struct pcs_int_request *ireq, struct pcs_msg *resp,
+ u32 io_lat, u32 net_lat)
{
struct pcs_cluster_core *cc = ireq->cc;
struct pcs_fuse_stat *stat = &cc->stat;
@@ -333,7 +502,13 @@ void pcs_fuse_stat_io_count(struct pcs_int_request *ireq, struct pcs_msg *resp)
se = req_stat_entry(&fstat->io, h->hdr.type);
if (likely(se)) {
+ u32 pending_lat = ktime_to_us(ktime_sub(ireq->ts_sent, ireq->ts));
+
spin_lock(&fstat->lock);
+ fuse_latency_count(&fstat->pending_lat, pending_lat);
+ fuse_latency_count(&fstat->io_lat, io_lat);
+ fuse_latency_count(&fstat->net_lat, net_lat);
+
fuse_val_stat_update(&se->curr, size);
spin_unlock(&fstat->lock);
}
@@ -436,6 +611,10 @@ void pcs_fuse_stat_init(struct pcs_fuse_stat *stat)
stat->fstat = fuse_kio_add_dentry(stat->kio_stat, fc, "fstat",
S_IFREG | S_IRUSR, 1, NULL,
&pcs_fuse_fstat_ops, stat);
+
+ stat->fstat_lat = fuse_kio_add_dentry(stat->kio_stat, fc, "fstat_lat",
+ S_IFREG | S_IRUSR, 1, NULL,
+ &pcs_fuse_fstat_lat_ops, stat);
out:
mutex_unlock(&fuse_mutex);
}
@@ -453,6 +632,8 @@ void pcs_fuse_stat_fini(struct pcs_fuse_stat *stat)
fuse_kio_rm_dentry(stat->requests);
if (stat->fstat)
fuse_kio_rm_dentry(stat->fstat);
+ if (stat->fstat_lat)
+ fuse_kio_rm_dentry(stat->fstat_lat);
fuse_kio_rm_dentry(stat->kio_stat);
}
mutex_unlock(&fuse_mutex);
diff --git a/fs/fuse/kio/pcs/fuse_stat.h b/fs/fuse/kio/pcs/fuse_stat.h
index b65620b51738..32971cf77dc4 100644
--- a/fs/fuse/kio/pcs/fuse_stat.h
+++ b/fs/fuse/kio/pcs/fuse_stat.h
@@ -34,11 +34,38 @@ struct pcs_fuse_stat {
struct dentry *iostat;
struct dentry *requests;
struct dentry *fstat;
+ struct dentry *fstat_lat;
+};
+
+enum {
+ LAT_ORDER1 = 0,
+ LAT_ORDER2 = 1,
+ LAT_ORDER3 = 2,
+ LAT_ORDER4 = 3,
+ LAT_ORDER5 = 4,
+ LAT_ORDER_OTHER = 5,
+};
+
+#define LATENCY_ORDER_MAX (LAT_ORDER_OTHER + 1)
+
+struct fuse_lat_stat {
+ u64 lat[LATENCY_ORDER_MAX];
+ u64 count;
+};
+
+struct fuse_lat_cnt {
+ struct fuse_lat_stat curr;
+ struct fuse_lat_stat last;
+ struct fuse_lat_stat glob;
};
struct fuse_io_cnt {
struct pcs_fuse_io_stat io;
+ struct fuse_lat_cnt io_lat;
+ struct fuse_lat_cnt net_lat;
+ struct fuse_lat_cnt pending_lat;
+
abs_time_t created_ts;
spinlock_t lock;
};
@@ -46,6 +73,7 @@ struct fuse_io_cnt {
void pcs_fuse_stat_init(struct pcs_fuse_stat *stat);
void pcs_fuse_stat_fini(struct pcs_fuse_stat *stat);
-void pcs_fuse_stat_io_count(struct pcs_int_request *ireq, struct pcs_msg *resp);
+void pcs_fuse_stat_io_count(struct pcs_int_request *ireq, struct pcs_msg *resp,
+ u32 io_lat, u32 net_lat);
#endif /* _FUSE_STAT_H_ */
diff --git a/fs/fuse/kio/pcs/pcs_map.c b/fs/fuse/kio/pcs/pcs_map.c
index 59de04be0512..c1cacae38fe4 100644
--- a/fs/fuse/kio/pcs/pcs_map.c
+++ b/fs/fuse/kio/pcs/pcs_map.c
@@ -2711,7 +2711,7 @@ static int commit_sync_info(struct pcs_int_request *req,
max_iolat = srec->sync.ts_io;
}
}
- pcs_fuse_stat_io_count(req, resp);
+ pcs_fuse_stat_io_count(req, resp, max_iolat, lat - max_iolat);
cs_log_io_times(req, resp, max_iolat);
evaluate_dirty_status(m);
More information about the Devel
mailing list