summaryrefslogtreecommitdiff
path: root/fs/ceph/metric.c
blob: 871c1090e520112d4a5641ab941bc1848b9a5e3d (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
/* SPDX-License-Identifier: GPL-2.0 */
#include <linux/ceph/ceph_debug.h>

#include <linux/types.h>
#include <linux/percpu_counter.h>
#include <linux/math64.h>

#include "metric.h"
#include "mds_client.h"

static void ktime_to_ceph_timespec(struct ceph_timespec *ts, ktime_t val)
{
	struct timespec64 t = ktime_to_timespec64(val);
	ceph_encode_timespec64(ts, &t);
}

static bool ceph_mdsc_send_metrics(struct ceph_mds_client *mdsc,
				   struct ceph_mds_session *s)
{
	struct ceph_metric_head *head;
	struct ceph_metric_cap *cap;
	struct ceph_metric_read_latency *read;
	struct ceph_metric_write_latency *write;
	struct ceph_metric_metadata_latency *meta;
	struct ceph_metric_dlease *dlease;
	struct ceph_opened_files *files;
	struct ceph_pinned_icaps *icaps;
	struct ceph_opened_inodes *inodes;
	struct ceph_read_io_size *rsize;
	struct ceph_write_io_size *wsize;
	struct ceph_client_metric *m = &mdsc->metric;
	u64 nr_caps = atomic64_read(&m->total_caps);
	u32 header_len = sizeof(struct ceph_metric_header);
	struct ceph_client *cl = mdsc->fsc->client;
	struct ceph_msg *msg;
	s64 sum;
	s32 items = 0;
	s32 len;

	/* Do not send the metrics until the MDS rank is ready */
	mutex_lock(&mdsc->mutex);
	if (ceph_mdsmap_get_state(mdsc->mdsmap, s->s_mds) != CEPH_MDS_STATE_ACTIVE) {
		mutex_unlock(&mdsc->mutex);
		return false;
	}
	mutex_unlock(&mdsc->mutex);

	len = sizeof(*head) + sizeof(*cap) + sizeof(*read) + sizeof(*write)
	      + sizeof(*meta) + sizeof(*dlease) + sizeof(*files)
	      + sizeof(*icaps) + sizeof(*inodes) + sizeof(*rsize)
	      + sizeof(*wsize);

	msg = ceph_msg_new(CEPH_MSG_CLIENT_METRICS, len, GFP_NOFS, true);
	if (!msg) {
		pr_err_client(cl, "to mds%d, failed to allocate message\n",
			      s->s_mds);
		return false;
	}

	head = msg->front.iov_base;

	/* encode the cap metric */
	cap = (struct ceph_metric_cap *)(head + 1);
	cap->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_CAP_INFO);
	cap->header.ver = 1;
	cap->header.compat = 1;
	cap->header.data_len = cpu_to_le32(sizeof(*cap) - header_len);
	cap->hit = cpu_to_le64(percpu_counter_sum(&m->i_caps_hit));
	cap->mis = cpu_to_le64(percpu_counter_sum(&m->i_caps_mis));
	cap->total = cpu_to_le64(nr_caps);
	items++;

	/* encode the read latency metric */
	read = (struct ceph_metric_read_latency *)(cap + 1);
	read->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_READ_LATENCY);
	read->header.ver = 2;
	read->header.compat = 1;
	read->header.data_len = cpu_to_le32(sizeof(*read) - header_len);
	sum = m->metric[METRIC_READ].latency_sum;
	ktime_to_ceph_timespec(&read->lat, sum);
	ktime_to_ceph_timespec(&read->avg, m->metric[METRIC_READ].latency_avg);
	read->sq_sum = cpu_to_le64(m->metric[METRIC_READ].latency_sq_sum);
	read->count = cpu_to_le64(m->metric[METRIC_READ].total);
	items++;

	/* encode the write latency metric */
	write = (struct ceph_metric_write_latency *)(read + 1);
	write->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_WRITE_LATENCY);
	write->header.ver = 2;
	write->header.compat = 1;
	write->header.data_len = cpu_to_le32(sizeof(*write) - header_len);
	sum = m->metric[METRIC_WRITE].latency_sum;
	ktime_to_ceph_timespec(&write->lat, sum);
	ktime_to_ceph_timespec(&write->avg, m->metric[METRIC_WRITE].latency_avg);
	write->sq_sum = cpu_to_le64(m->metric[METRIC_WRITE].latency_sq_sum);
	write->count = cpu_to_le64(m->metric[METRIC_WRITE].total);
	items++;

	/* encode the metadata latency metric */
	meta = (struct ceph_metric_metadata_latency *)(write + 1);
	meta->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_METADATA_LATENCY);
	meta->header.ver = 2;
	meta->header.compat = 1;
	meta->header.data_len = cpu_to_le32(sizeof(*meta) - header_len);
	sum = m->metric[METRIC_METADATA].latency_sum;
	ktime_to_ceph_timespec(&meta->lat, sum);
	ktime_to_ceph_timespec(&meta->avg, m->metric[METRIC_METADATA].latency_avg);
	meta->sq_sum = cpu_to_le64(m->metric[METRIC_METADATA].latency_sq_sum);
	meta->count = cpu_to_le64(m->metric[METRIC_METADATA].total);
	items++;

	/* encode the dentry lease metric */
	dlease = (struct ceph_metric_dlease *)(meta + 1);
	dlease->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_DENTRY_LEASE);
	dlease->header.ver = 1;
	dlease->header.compat = 1;
	dlease->header.data_len = cpu_to_le32(sizeof(*dlease) - header_len);
	dlease->hit = cpu_to_le64(percpu_counter_sum(&m->d_lease_hit));
	dlease->mis = cpu_to_le64(percpu_counter_sum(&m->d_lease_mis));
	dlease->total = cpu_to_le64(atomic64_read(&m->total_dentries));
	items++;

	sum = percpu_counter_sum(&m->total_inodes);

	/* encode the opened files metric */
	files = (struct ceph_opened_files *)(dlease + 1);
	files->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_OPENED_FILES);
	files->header.ver = 1;
	files->header.compat = 1;
	files->header.data_len = cpu_to_le32(sizeof(*files) - header_len);
	files->opened_files = cpu_to_le64(atomic64_read(&m->opened_files));
	files->total = cpu_to_le64(sum);
	items++;

	/* encode the pinned icaps metric */
	icaps = (struct ceph_pinned_icaps *)(files + 1);
	icaps->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_PINNED_ICAPS);
	icaps->header.ver = 1;
	icaps->header.compat = 1;
	icaps->header.data_len = cpu_to_le32(sizeof(*icaps) - header_len);
	icaps->pinned_icaps = cpu_to_le64(nr_caps);
	icaps->total = cpu_to_le64(sum);
	items++;

	/* encode the opened inodes metric */
	inodes = (struct ceph_opened_inodes *)(icaps + 1);
	inodes->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_OPENED_INODES);
	inodes->header.ver = 1;
	inodes->header.compat = 1;
	inodes->header.data_len = cpu_to_le32(sizeof(*inodes) - header_len);
	inodes->opened_inodes = cpu_to_le64(percpu_counter_sum(&m->opened_inodes));
	inodes->total = cpu_to_le64(sum);
	items++;

	/* encode the read io size metric */
	rsize = (struct ceph_read_io_size *)(inodes + 1);
	rsize->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_READ_IO_SIZES);
	rsize->header.ver = 1;
	rsize->header.compat = 1;
	rsize->header.data_len = cpu_to_le32(sizeof(*rsize) - header_len);
	rsize->total_ops = cpu_to_le64(m->metric[METRIC_READ].total);
	rsize->total_size = cpu_to_le64(m->metric[METRIC_READ].size_sum);
	items++;

	/* encode the write io size metric */
	wsize = (struct ceph_write_io_size *)(rsize + 1);
	wsize->header.type = cpu_to_le32(CLIENT_METRIC_TYPE_WRITE_IO_SIZES);
	wsize->header.ver = 1;
	wsize->header.compat = 1;
	wsize->header.data_len = cpu_to_le32(sizeof(*wsize) - header_len);
	wsize->total_ops = cpu_to_le64(m->metric[METRIC_WRITE].total);
	wsize->total_size = cpu_to_le64(m->metric[METRIC_WRITE].size_sum);
	items++;

	put_unaligned_le32(items, &head->num);
	msg->front.iov_len = len;
	msg->hdr.version = cpu_to_le16(1);
	msg->hdr.compat_version = cpu_to_le16(1);
	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
	ceph_con_send(&s->s_con, msg);

	return true;
}


static void metric_get_session(struct ceph_mds_client *mdsc)
{
	struct ceph_mds_session *s;
	int i;

	mutex_lock(&mdsc->mutex);
	for (i = 0; i < mdsc->max_sessions; i++) {
		s = __ceph_lookup_mds_session(mdsc, i);
		if (!s)
			continue;

		/*
		 * Skip it if MDS doesn't support the metric collection,
		 * or the MDS will close the session's socket connection
		 * directly when it get this message.
		 */
		if (check_session_state(s) &&
		    test_bit(CEPHFS_FEATURE_METRIC_COLLECT, &s->s_features)) {
			mdsc->metric.session = s;
			break;
		}

		ceph_put_mds_session(s);
	}
	mutex_unlock(&mdsc->mutex);
}

static void metric_delayed_work(struct work_struct *work)
{
	struct ceph_client_metric *m =
		container_of(work, struct ceph_client_metric, delayed_work.work);
	struct ceph_mds_client *mdsc =
		container_of(m, struct ceph_mds_client, metric);

	if (mdsc->stopping || disable_send_metrics)
		return;

	if (!m->session || !check_session_state(m->session)) {
		if (m->session) {
			ceph_put_mds_session(m->session);
			m->session = NULL;
		}
		metric_get_session(mdsc);
	}
	if (m->session) {
		ceph_mdsc_send_metrics(mdsc, m->session);
		metric_schedule_delayed(m);
	}
}

int ceph_metric_init(struct ceph_client_metric *m)
{
	struct ceph_metric *metric;
	int ret, i;

	if (!m)
		return -EINVAL;

	atomic64_set(&m->total_dentries, 0);
	ret = percpu_counter_init(&m->d_lease_hit, 0, GFP_KERNEL);
	if (ret)
		return ret;

	ret = percpu_counter_init(&m->d_lease_mis, 0, GFP_KERNEL);
	if (ret)
		goto err_d_lease_mis;

	atomic64_set(&m->total_caps, 0);
	ret = percpu_counter_init(&m->i_caps_hit, 0, GFP_KERNEL);
	if (ret)
		goto err_i_caps_hit;

	ret = percpu_counter_init(&m->i_caps_mis, 0, GFP_KERNEL);
	if (ret)
		goto err_i_caps_mis;

	for (i = 0; i < METRIC_MAX; i++) {
		metric = &m->metric[i];
		spin_lock_init(&metric->lock);
		metric->size_sum = 0;
		metric->size_min = U64_MAX;
		metric->size_max = 0;
		metric->total = 0;
		metric->latency_sum = 0;
		metric->latency_avg = 0;
		metric->latency_sq_sum = 0;
		metric->latency_min = KTIME_MAX;
		metric->latency_max = 0;
	}

	atomic64_set(&m->opened_files, 0);
	ret = percpu_counter_init(&m->opened_inodes, 0, GFP_KERNEL);
	if (ret)
		goto err_opened_inodes;
	ret = percpu_counter_init(&m->total_inodes, 0, GFP_KERNEL);
	if (ret)
		goto err_total_inodes;

	m->session = NULL;
	INIT_DELAYED_WORK(&m->delayed_work, metric_delayed_work);

	return 0;

err_total_inodes:
	percpu_counter_destroy(&m->opened_inodes);
err_opened_inodes:
	percpu_counter_destroy(&m->i_caps_mis);
err_i_caps_mis:
	percpu_counter_destroy(&m->i_caps_hit);
err_i_caps_hit:
	percpu_counter_destroy(&m->d_lease_mis);
err_d_lease_mis:
	percpu_counter_destroy(&m->d_lease_hit);

	return ret;
}

void ceph_metric_destroy(struct ceph_client_metric *m)
{
	if (!m)
		return;

	cancel_delayed_work_sync(&m->delayed_work);

	percpu_counter_destroy(&m->total_inodes);
	percpu_counter_destroy(&m->opened_inodes);
	percpu_counter_destroy(&m->i_caps_mis);
	percpu_counter_destroy(&m->i_caps_hit);
	percpu_counter_destroy(&m->d_lease_mis);
	percpu_counter_destroy(&m->d_lease_hit);

	ceph_put_mds_session(m->session);
}

#define METRIC_UPDATE_MIN_MAX(min, max, new)	\
{						\
	if (unlikely(new < min))		\
		min = new;			\
	if (unlikely(new > max))		\
		max = new;			\
}

static inline void __update_mean_and_stdev(ktime_t total, ktime_t *lavg,
					   ktime_t *sq_sump, ktime_t lat)
{
	ktime_t avg;

	if (unlikely(total == 1)) {
		*lavg = lat;
	} else {
		/* the sq is (lat - old_avg) * (lat - new_avg) */
		avg = *lavg + div64_s64(lat - *lavg, total);
		*sq_sump += (lat - *lavg)*(lat - avg);
		*lavg = avg;
	}
}

void ceph_update_metrics(struct ceph_metric *m,
			 ktime_t r_start, ktime_t r_end,
			 unsigned int size, int rc)
{
	ktime_t lat = ktime_sub(r_end, r_start);
	ktime_t total;

	if (unlikely(rc < 0 && rc != -ENOENT && rc != -ETIMEDOUT))
		return;

	spin_lock(&m->lock);
	total = ++m->total;
	m->size_sum += size;
	METRIC_UPDATE_MIN_MAX(m->size_min, m->size_max, size);
	m->latency_sum += lat;
	METRIC_UPDATE_MIN_MAX(m->latency_min, m->latency_max, lat);
	__update_mean_and_stdev(total, &m->latency_avg,	&m->latency_sq_sum,
				lat);
	spin_unlock(&m->lock);
}