[v7,4/5] ceph: add global write latency metric support

Commit Message

Xiubo Li Feb. 19, 2020, 3:38 a.m. UTC

From: Xiubo Li <[e-mail]>

It will calculate the latency for the write osd requests, which only
include the time cousumed by network and the ceph osd.

item          total       sum_lat(us)     avg_lat(us)
-----------------------------------------------------
write         1048        8778000         8375

URL: https://tracker.ceph.com/issues/43215
Signed-off-by: Xiubo Li <[e-mail]>
---
 fs/ceph/addr.c       |  7 +++++++
 fs/ceph/debugfs.c    |  6 ++++++
 fs/ceph/file.c       | 13 ++++++++++---
 fs/ceph/mds_client.c | 14 ++++++++++++++
 fs/ceph/metric.h     | 18 ++++++++++++++++++
 5 files changed, 55 insertions(+), 3 deletions(-)

Patch

diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c
index 16573a13ffee..aca2ca592e53 100644
--- a/fs/ceph/addr.c
+++ b/fs/ceph/addr.c
@@ -649,6 +649,8 @@  static int ceph_sync_writepages(struct ceph_fs_client *fsc,
 	if (!rc)
 		rc = ceph_osdc_wait_request(osdc, req);
 
+	ceph_update_write_latency(&fsc->mdsc->metric, req, rc);
+
 	ceph_osdc_put_request(req);
 	if (rc == 0)
 		rc = len;
@@ -800,6 +802,8 @@  static void writepages_finish(struct ceph_osd_request *req)
 		ceph_clear_error_write(ci);
 	}
 
+	ceph_update_write_latency(&fsc->mdsc->metric, req, rc);
+
 	/*
 	 * We lost the cache cap, need to truncate the page before
 	 * it is unlocked, otherwise we'd truncate it later in the
@@ -1858,6 +1862,9 @@  int ceph_uninline_data(struct file *filp, struct page *locked_page)
 	err = ceph_osdc_start_request(&fsc->client->osdc, req, false);
 	if (!err)
 		err = ceph_osdc_wait_request(&fsc->client->osdc, req);
+
+	ceph_update_write_latency(&fsc->mdsc->metric, req, err);
+
 out_put:
 	ceph_osdc_put_request(req);
 	if (err == -ECANCELED)
diff --git a/fs/ceph/debugfs.c b/fs/ceph/debugfs.c
index d814a3a27611..464bfbdb970d 100644
--- a/fs/ceph/debugfs.c
+++ b/fs/ceph/debugfs.c
@@ -140,6 +140,12 @@  static int metric_show(struct seq_file *s, void *p)
 	avg = total ? sum / total : 0;
 	seq_printf(s, "%-14s%-12lld%-16lld%lld\n", "read", total, sum, avg);
 
+	total = percpu_counter_sum(&mdsc->metric.total_writes);
+	sum = percpu_counter_sum(&mdsc->metric.write_latency_sum);
+	sum = jiffies_to_usecs(sum);
+	avg = total ? sum / total : 0;
+	seq_printf(s, "%-14s%-12lld%-16lld%lld\n", "write", total, sum, avg);
+
 	seq_printf(s, "\n");
 	seq_printf(s, "item          total           miss            hit\n");
 	seq_printf(s, "-------------------------------------------------\n");
diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index 96e35935b764..0a25dc7e3a52 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -811,8 +811,12 @@  static void ceph_aio_complete_req(struct ceph_osd_request *req)
 	     inode, rc, osd_data->bvec_pos.iter.bi_size);
 
 	/* r_start_stamp == 0 means the request was not submitted */
-	if (req->r_start_stamp && !aio_req->write)
-		ceph_update_read_latency(metric, req, rc);
+	if (req->r_start_stamp) {
+		if (aio_req->write)
+			ceph_update_write_latency(metric, req, rc);
+		else
+			ceph_update_read_latency(metric, req, rc);
+	}
 
 	if (rc == -EOLDSNAPC) {
 		struct ceph_aio_work *aio_work;
@@ -1059,7 +1063,9 @@  ceph_direct_read_write(struct kiocb *iocb, struct iov_iter *iter,
 		if (!ret)
 			ret = ceph_osdc_wait_request(&fsc->client->osdc, req);
 
-		if (!write)
+		if (write)
+			ceph_update_write_latency(metric, req, ret);
+		else
 			ceph_update_read_latency(metric, req, ret);
 
 		size = i_size_read(inode);
@@ -1233,6 +1239,7 @@  ceph_sync_write(struct kiocb *iocb, struct iov_iter *from, loff_t pos,
 		if (!ret)
 			ret = ceph_osdc_wait_request(&fsc->client->osdc, req);
 
+		ceph_update_write_latency(&fsc->mdsc->metric, req, ret);
 out:
 		ceph_osdc_put_request(req);
 		if (ret != 0) {
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index b7ada9cde4f8..58e97ac004d6 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -4192,8 +4192,20 @@  static int ceph_mdsc_metric_init(struct ceph_client_metric *metric)
 	if (ret)
 		goto err_read_latency_sum;
 
+	ret = percpu_counter_init(&metric->total_writes, 0, GFP_KERNEL);
+	if (ret)
+		goto err_total_writes;
+
+	ret = percpu_counter_init(&metric->write_latency_sum, 0, GFP_KERNEL);
+	if (ret)
+		goto err_write_latency_sum;
+
 	return 0;
 
+err_write_latency_sum:
+	percpu_counter_destroy(&metric->total_writes);
+err_total_writes:
+	percpu_counter_destroy(&metric->read_latency_sum);
 err_read_latency_sum:
 	percpu_counter_destroy(&metric->total_reads);
 err_total_reads:
@@ -4545,6 +4557,8 @@  void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
 
 	ceph_mdsc_stop(mdsc);
 
+	percpu_counter_destroy(&mdsc->metric.write_latency_sum);
+	percpu_counter_destroy(&mdsc->metric.total_writes);
 	percpu_counter_destroy(&mdsc->metric.read_latency_sum);
 	percpu_counter_destroy(&mdsc->metric.total_reads);
 	percpu_counter_destroy(&mdsc->metric.i_caps_mis);
diff --git a/fs/ceph/metric.h b/fs/ceph/metric.h
index afea44a3794b..a87197f3e915 100644
--- a/fs/ceph/metric.h
+++ b/fs/ceph/metric.h
@@ -15,6 +15,9 @@  struct ceph_client_metric {
 
 	struct percpu_counter total_reads;
 	struct percpu_counter read_latency_sum;
+
+	struct percpu_counter total_writes;
+	struct percpu_counter write_latency_sum;
 };
 
 static inline void ceph_update_read_latency(struct ceph_client_metric *m,
@@ -31,4 +34,19 @@  static inline void ceph_update_read_latency(struct ceph_client_metric *m,
 		percpu_counter_add(&m->read_latency_sum, latency);
 	}
 }
+
+static inline void ceph_update_write_latency(struct ceph_client_metric *m,
+					     struct ceph_osd_request *req,
+					     int rc)
+{
+	if (!m || !req)
+		return;
+
+	if (!rc || rc == -ETIMEDOUT) {
+		s64 latency = req->r_end_stamp - req->r_start_stamp;
+
+		percpu_counter_inc(&m->total_writes);
+		percpu_counter_add(&m->write_latency_sum, latency);
+	}
+}
 #endif /* _FS_CEPH_MDS_METRIC_H */