summaryrefslogtreecommitdiffstats
path: root/libglusterfs/src/monitoring.c
blob: fbb68dc862275dc8583449999a08cbef87a3e8c3 (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
/*
  Copyright (c) 2017 Red Hat, Inc. <http://www.redhat.com>
  This file is part of GlusterFS.

  This file is licensed to you under your choice of the GNU Lesser
  General Public License, version 3 or any later version (LGPLv3 or
  later), or the GNU General Public License, version 2 (GPLv2), in all
  cases as published by the Free Software Foundation.
*/

#include "glusterfs/monitoring.h"
#include "glusterfs/xlator.h"
#include "glusterfs/syscall.h"

#include <stdlib.h>

static void
dump_mem_acct_details(xlator_t *xl, int fd)
{
    struct mem_acct_rec *mem_rec;
    int i = 0;

    if (!xl || !xl->mem_acct || (xl->ctx->active != xl->graph))
        return;

    dprintf(fd, "# %s.%s.total.num_types %d\n", xl->type, xl->name,
            xl->mem_acct->num_types);

    dprintf(fd,
            "# type, in-use-size, in-use-units, max-size, "
            "max-units, total-allocs\n");

    for (i = 0; i < xl->mem_acct->num_types; i++) {
        mem_rec = &xl->mem_acct->rec[i];
        if (mem_rec->num_allocs == 0)
            continue;
        dprintf(fd, "# %s, %" PRIu64 ", %u, %" PRIu64 ", %u, %" PRIu64 "\n",
                mem_rec->typestr, mem_rec->size, mem_rec->num_allocs,
                mem_rec->max_size, mem_rec->max_num_allocs,
                mem_rec->total_allocs);
    }
}

static void
dump_global_memory_accounting(int fd)
{
#if MEMORY_ACCOUNTING_STATS
    int i = 0;
    uint64_t count = 0;

    uint64_t tcalloc = GF_ATOMIC_GET(gf_memory_stat_counts.total_calloc);
    uint64_t tmalloc = GF_ATOMIC_GET(gf_memory_stat_counts.total_malloc);
    uint64_t tfree = GF_ATOMIC_GET(gf_memory_stat_counts.total_free);

    dprintf(fd, "memory.total.calloc %lu\n", tcalloc);
    dprintf(fd, "memory.total.malloc %lu\n", tmalloc);
    dprintf(fd, "memory.total.realloc %lu\n",
            GF_ATOMIC_GET(gf_memory_stat_counts.total_realloc));
    dprintf(fd, "memory.total.free %lu\n", tfree);
    dprintf(fd, "memory.total.in-use %lu\n", ((tcalloc + tmalloc) - tfree));

    for (i = 0; i < GF_BLK_MAX_VALUE; i++) {
        count = GF_ATOMIC_GET(gf_memory_stat_counts.blk_size[i]);
        dprintf(fd, "memory.total.blk_size.%s %lu\n",
                gf_mem_stats_blk[i].blk_size_str, count);
    }

    dprintf(fd, "#----\n");
#endif

    /* This is not a metric to be watched in admin guide,
       but keeping it here till we resolve all leak-issues
       would be great */
}

static void
dump_latency_and_count(xlator_t *xl, int fd)
{
    int32_t index = 0;
    uint64_t fop;
    uint64_t cbk;
    uint64_t count;

    if (xl->winds) {
        dprintf(fd, "%s.total.pending-winds.count %" PRIu64 "\n", xl->name,
                xl->winds);
    }

    /* Need 'fuse' data, and don't need all the old graph info */
    if ((xl != xl->ctx->master) && (xl->ctx->active != xl->graph))
        return;

    count = GF_ATOMIC_GET(xl->stats.total.count);
    dprintf(fd, "%s.total.fop-count %" PRIu64 "\n", xl->name, count);

    count = GF_ATOMIC_GET(xl->stats.interval.count);
    dprintf(fd, "%s.interval.fop-count %" PRIu64 "\n", xl->name, count);
    GF_ATOMIC_INIT(xl->stats.interval.count, 0);

    for (index = 0; index < GF_FOP_MAXVALUE; index++) {
        fop = GF_ATOMIC_GET(xl->stats.total.metrics[index].fop);
        if (fop) {
            dprintf(fd, "%s.total.%s.count %" PRIu64 "\n", xl->name,
                    gf_fop_list[index], fop);
        }
        fop = GF_ATOMIC_GET(xl->stats.interval.metrics[index].fop);
        if (fop) {
            dprintf(fd, "%s.interval.%s.count %" PRIu64 "\n", xl->name,
                    gf_fop_list[index], fop);
        }
        cbk = GF_ATOMIC_GET(xl->stats.interval.metrics[index].cbk);
        if (cbk) {
            dprintf(fd, "%s.interval.%s.fail_count %" PRIu64 "\n", xl->name,
                    gf_fop_list[index], cbk);
        }
        if (xl->stats.interval.latencies[index].count != 0) {
            dprintf(fd, "%s.interval.%s.latency %lf\n", xl->name,
                    gf_fop_list[index],
                    (((double)xl->stats.interval.latencies[index].total) /
                     xl->stats.interval.latencies[index].count));
            dprintf(fd, "%s.interval.%s.max %" PRIu64 "\n", xl->name,
                    gf_fop_list[index],
                    xl->stats.interval.latencies[index].max);
            dprintf(fd, "%s.interval.%s.min %" PRIu64 "\n", xl->name,
                    gf_fop_list[index],
                    xl->stats.interval.latencies[index].min);
        }
        GF_ATOMIC_INIT(xl->stats.interval.metrics[index].cbk, 0);
        GF_ATOMIC_INIT(xl->stats.interval.metrics[index].fop, 0);
    }
    memset(xl->stats.interval.latencies, 0,
           sizeof(xl->stats.interval.latencies));
}

static inline void
dump_call_stack_details(glusterfs_ctx_t *ctx, int fd)
{
    dprintf(fd, "total.stack.count %" PRIu64 "\n",
            GF_ATOMIC_GET(ctx->pool->total_count));
    dprintf(fd, "total.stack.in-flight %" PRIu64 "\n", ctx->pool->cnt);
}

static inline void
dump_dict_details(glusterfs_ctx_t *ctx, int fd)
{
    uint64_t total_dicts = 0;
    uint64_t total_pairs = 0;

    total_dicts = GF_ATOMIC_GET(ctx->stats.total_dicts_used);
    total_pairs = GF_ATOMIC_GET(ctx->stats.total_pairs_used);

    dprintf(fd, "total.dict.max-pairs-per %" PRIu64 "\n",
            GF_ATOMIC_GET(ctx->stats.max_dict_pairs));
    dprintf(fd, "total.dict.pairs-used %" PRIu64 "\n", total_pairs);
    dprintf(fd, "total.dict.used %" PRIu64 "\n", total_dicts);
    dprintf(fd, "total.dict.average-pairs %" PRIu64 "\n",
            (total_pairs / total_dicts));
}

static void
dump_inode_stats(glusterfs_ctx_t *ctx, int fd)
{
}

static void
dump_global_metrics(glusterfs_ctx_t *ctx, int fd)
{
    struct timeval tv;
    time_t nowtime;
    struct tm *nowtm;
    char tmbuf[64] = {
        0,
    };

    gettimeofday(&tv, NULL);
    nowtime = tv.tv_sec;
    nowtm = localtime(&nowtime);
    strftime(tmbuf, sizeof tmbuf, "%Y-%m-%d %H:%M:%S", nowtm);

    /* Let every file have information on which process dumped info */
    dprintf(fd, "## %s\n", ctx->cmdlinestr);
    dprintf(fd, "### %s\n", tmbuf);
    dprintf(fd, "### BrickName: %s\n", ctx->cmd_args.brick_name);
    dprintf(fd, "### MountName: %s\n", ctx->cmd_args.mount_point);
    dprintf(fd, "### VolumeName: %s\n", ctx->cmd_args.volume_name);

    /* Dump memory accounting */
    dump_global_memory_accounting(fd);
    dprintf(fd, "# -----\n");

    dump_call_stack_details(ctx, fd);
    dump_dict_details(ctx, fd);
    dprintf(fd, "# -----\n");

    dump_inode_stats(ctx, fd);
    dprintf(fd, "# -----\n");
}

static void
dump_xl_metrics(glusterfs_ctx_t *ctx, int fd)
{
    xlator_t *xl;

    xl = ctx->active->top;

    while (xl) {
        dump_latency_and_count(xl, fd);
        dump_mem_acct_details(xl, fd);
        if (xl->dump_metrics)
            xl->dump_metrics(xl, fd);
        xl = xl->next;
    }

    if (ctx->master) {
        xl = ctx->master;

        dump_latency_and_count(xl, fd);
        dump_mem_acct_details(xl, fd);
        if (xl->dump_metrics)
            xl->dump_metrics(xl, fd);
    }

    return;
}

char *
gf_monitor_metrics(glusterfs_ctx_t *ctx)
{
    int ret = -1;
    int fd = 0;
    char *filepath = NULL, *dumppath = NULL;

    gf_msg_trace("monitoring", 0, "received monitoring request (sig:USR2)");

    dumppath = ctx->config.metrics_dumppath;
    if (dumppath == NULL) {
        dumppath = GLUSTER_METRICS_DIR;
    }
    ret = mkdir_p(dumppath, 0755, true);
    if (ret) {
        /* EEXIST is handled in mkdir_p() itself */
        gf_msg("monitoring", GF_LOG_ERROR, 0, LG_MSG_STRDUP_ERROR,
               "failed to create metrics dir %s (%s)", dumppath,
               strerror(errno));
        return NULL;
    }

    ret = gf_asprintf(&filepath, "%s/gmetrics.XXXXXX", dumppath);
    if (ret < 0) {
        return NULL;
    }

    /* coverity[secure_temp] mkstemp uses 0600 as the mode and is safe */
    fd = mkstemp(filepath);
    if (fd < 0) {
        gf_msg("monitoring", GF_LOG_ERROR, 0, LG_MSG_STRDUP_ERROR,
               "failed to open tmp file %s (%s)", filepath, strerror(errno));
        GF_FREE(filepath);
        return NULL;
    }

    dump_global_metrics(ctx, fd);

    dump_xl_metrics(ctx, fd);

    /* This below line is used just to capture any errors with dprintf() */
    ret = dprintf(fd, "\n# End of metrics\n");
    if (ret < 0) {
        gf_msg("monitoring", GF_LOG_WARNING, 0, LG_MSG_STRDUP_ERROR,
               "dprintf() failed: %s", strerror(errno));
    }

    ret = sys_fsync(fd);
    if (ret < 0) {
        gf_msg("monitoring", GF_LOG_WARNING, 0, LG_MSG_STRDUP_ERROR,
               "fsync() failed: %s", strerror(errno));
    }
    sys_close(fd);

    /* Figure this out, not happy with returning this string */
    return filepath;
}