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
|
// -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
// vim: ts=8 sw=2 smarttab
#include "osd_perf_counters.h"
#include "include/common_fwd.h"
PerfCounters *build_osd_logger(CephContext *cct) {
PerfCountersBuilder osd_plb(cct, "osd", l_osd_first, l_osd_last);
// Latency axis configuration for op histograms, values are in nanoseconds
PerfHistogramCommon::axis_config_d op_hist_x_axis_config{
"Latency (usec)",
PerfHistogramCommon::SCALE_LOG2, ///< Latency in logarithmic scale
0, ///< Start at 0
100000, ///< Quantization unit is 100usec
32, ///< Enough to cover much longer than slow requests
};
// Op size axis configuration for op histograms, values are in bytes
PerfHistogramCommon::axis_config_d op_hist_y_axis_config{
"Request size (bytes)",
PerfHistogramCommon::SCALE_LOG2, ///< Request size in logarithmic scale
0, ///< Start at 0
512, ///< Quantization unit is 512 bytes
32, ///< Enough to cover requests larger than GB
};
// All the basic OSD operation stats are to be considered useful
osd_plb.set_prio_default(PerfCountersBuilder::PRIO_USEFUL);
osd_plb.add_u64(
l_osd_op_wip, "op_wip",
"Replication operations currently being processed (primary)");
osd_plb.add_u64_counter(
l_osd_op, "op",
"Client operations",
"ops", PerfCountersBuilder::PRIO_CRITICAL);
osd_plb.add_u64_counter(
l_osd_op_inb, "op_in_bytes",
"Client operations total write size",
"wr", PerfCountersBuilder::PRIO_INTERESTING, unit_t(UNIT_BYTES));
osd_plb.add_u64_counter(
l_osd_op_outb, "op_out_bytes",
"Client operations total read size",
"rd", PerfCountersBuilder::PRIO_INTERESTING, unit_t(UNIT_BYTES));
osd_plb.add_time_avg(
l_osd_op_lat, "op_latency",
"Latency of client operations (including queue time)",
"l", 9);
osd_plb.add_time_avg(
l_osd_op_process_lat, "op_process_latency",
"Latency of client operations (excluding queue time)");
osd_plb.add_time_avg(
l_osd_op_prepare_lat, "op_prepare_latency",
"Latency of client operations (excluding queue time and wait for finished)");
osd_plb.add_u64_counter(
l_osd_op_delayed_unreadable, "op_delayed_unreadable",
"Count of ops delayed due to target object being unreadable");
osd_plb.add_u64_counter(
l_osd_op_delayed_degraded, "op_delayed_degraded",
"Count of ops delayed due to target object being degraded");
osd_plb.add_u64_counter(
l_osd_op_r, "op_r", "Client read operations");
osd_plb.add_u64_counter(
l_osd_op_r_outb, "op_r_out_bytes", "Client data read", NULL, PerfCountersBuilder::PRIO_USEFUL, unit_t(UNIT_BYTES));
osd_plb.add_time_avg(
l_osd_op_r_lat, "op_r_latency",
"Latency of read operation (including queue time)");
osd_plb.add_u64_counter_histogram(
l_osd_op_r_lat_outb_hist, "op_r_latency_out_bytes_histogram",
op_hist_x_axis_config, op_hist_y_axis_config,
"Histogram of operation latency (including queue time) + data read");
osd_plb.add_time_avg(
l_osd_op_r_process_lat, "op_r_process_latency",
"Latency of read operation (excluding queue time)");
osd_plb.add_time_avg(
l_osd_op_r_prepare_lat, "op_r_prepare_latency",
"Latency of read operations (excluding queue time and wait for finished)");
osd_plb.add_u64_counter(
l_osd_op_w, "op_w", "Client write operations");
osd_plb.add_u64_counter(
l_osd_op_w_inb, "op_w_in_bytes", "Client data written");
osd_plb.add_time_avg(
l_osd_op_w_lat, "op_w_latency",
"Latency of write operation (including queue time)");
osd_plb.add_u64_counter_histogram(
l_osd_op_w_lat_inb_hist, "op_w_latency_in_bytes_histogram",
op_hist_x_axis_config, op_hist_y_axis_config,
"Histogram of operation latency (including queue time) + data written");
osd_plb.add_time_avg(
l_osd_op_w_process_lat, "op_w_process_latency",
"Latency of write operation (excluding queue time)");
osd_plb.add_time_avg(
l_osd_op_w_prepare_lat, "op_w_prepare_latency",
"Latency of write operations (excluding queue time and wait for finished)");
osd_plb.add_u64_counter(
l_osd_op_rw, "op_rw",
"Client read-modify-write operations");
osd_plb.add_u64_counter(
l_osd_op_rw_inb, "op_rw_in_bytes",
"Client read-modify-write operations write in", NULL, PerfCountersBuilder::PRIO_USEFUL, unit_t(UNIT_BYTES));
osd_plb.add_u64_counter(
l_osd_op_rw_outb,"op_rw_out_bytes",
"Client read-modify-write operations read out ", NULL, PerfCountersBuilder::PRIO_USEFUL, unit_t(UNIT_BYTES));
osd_plb.add_time_avg(
l_osd_op_rw_lat, "op_rw_latency",
"Latency of read-modify-write operation (including queue time)");
osd_plb.add_u64_counter_histogram(
l_osd_op_rw_lat_inb_hist, "op_rw_latency_in_bytes_histogram",
op_hist_x_axis_config, op_hist_y_axis_config,
"Histogram of rw operation latency (including queue time) + data written");
osd_plb.add_u64_counter_histogram(
l_osd_op_rw_lat_outb_hist, "op_rw_latency_out_bytes_histogram",
op_hist_x_axis_config, op_hist_y_axis_config,
"Histogram of rw operation latency (including queue time) + data read");
osd_plb.add_time_avg(
l_osd_op_rw_process_lat, "op_rw_process_latency",
"Latency of read-modify-write operation (excluding queue time)");
osd_plb.add_time_avg(
l_osd_op_rw_prepare_lat, "op_rw_prepare_latency",
"Latency of read-modify-write operations (excluding queue time and wait for finished)");
// Now we move on to some more obscure stats, revert to assuming things
// are low priority unless otherwise specified.
osd_plb.set_prio_default(PerfCountersBuilder::PRIO_DEBUGONLY);
osd_plb.add_time_avg(l_osd_op_before_queue_op_lat, "op_before_queue_op_lat",
"Latency of IO before calling queue(before really queue into ShardedOpWq)"); // client io before queue op_wq latency
osd_plb.add_time_avg(l_osd_op_before_dequeue_op_lat, "op_before_dequeue_op_lat",
"Latency of IO before calling dequeue_op(already dequeued and get PG lock)"); // client io before dequeue_op latency
osd_plb.add_u64_counter(
l_osd_sop, "subop", "Suboperations");
osd_plb.add_u64_counter(
l_osd_sop_inb, "subop_in_bytes", "Suboperations total size", NULL, 0, unit_t(UNIT_BYTES));
osd_plb.add_time_avg(l_osd_sop_lat, "subop_latency", "Suboperations latency");
osd_plb.add_u64_counter(l_osd_sop_w, "subop_w", "Replicated writes");
osd_plb.add_u64_counter(
l_osd_sop_w_inb, "subop_w_in_bytes", "Replicated written data size", NULL, 0, unit_t(UNIT_BYTES));
osd_plb.add_time_avg(
l_osd_sop_w_lat, "subop_w_latency", "Replicated writes latency");
osd_plb.add_u64_counter(
l_osd_sop_pull, "subop_pull", "Suboperations pull requests");
osd_plb.add_time_avg(
l_osd_sop_pull_lat, "subop_pull_latency", "Suboperations pull latency");
osd_plb.add_u64_counter(
l_osd_sop_push, "subop_push", "Suboperations push messages");
osd_plb.add_u64_counter(
l_osd_sop_push_inb, "subop_push_in_bytes", "Suboperations pushed size", NULL, 0, unit_t(UNIT_BYTES));
osd_plb.add_time_avg(
l_osd_sop_push_lat, "subop_push_latency", "Suboperations push latency");
osd_plb.add_u64_counter(l_osd_pull, "pull", "Pull requests sent");
osd_plb.add_u64_counter(l_osd_push, "push", "Push messages sent");
osd_plb.add_u64_counter(l_osd_push_outb, "push_out_bytes", "Pushed size", NULL, 0, unit_t(UNIT_BYTES));
osd_plb.add_u64_counter(
l_osd_rop, "recovery_ops",
"Started recovery operations",
"rop", PerfCountersBuilder::PRIO_INTERESTING);
osd_plb.add_u64_counter(
l_osd_rbytes, "recovery_bytes",
"recovery bytes",
"rbt", PerfCountersBuilder::PRIO_INTERESTING);
osd_plb.add_time_avg(
l_osd_recovery_push_queue_lat,
"l_osd_recovery_push_queue_latency",
"MOSDPGPush queue latency");
osd_plb.add_time_avg(
l_osd_recovery_push_reply_queue_lat,
"l_osd_recovery_push_reply_queue_latency",
"MOSDPGPushReply queue latency");
osd_plb.add_time_avg(
l_osd_recovery_pull_queue_lat,
"l_osd_recovery_pull_queue_latency",
"MOSDPGPull queue latency");
osd_plb.add_time_avg(
l_osd_recovery_backfill_queue_lat,
"l_osd_recovery_backfill_queue_latency",
"MOSDPGBackfill queue latency");
osd_plb.add_time_avg(
l_osd_recovery_backfill_remove_queue_lat,
"l_osd_recovery_backfill_remove_queue_latency",
"MOSDPGBackfillDelete queue latency");
osd_plb.add_time_avg(
l_osd_recovery_scan_queue_lat,
"l_osd_recovery_scan_queue_latency",
"MOSDPGScan queue latency");
osd_plb.add_time_avg(
l_osd_recovery_queue_lat,
"l_osd_recovery_queue_latency",
"PGRecovery queue latency");
osd_plb.add_time_avg(
l_osd_recovery_context_queue_lat,
"l_osd_recovery_context_queue_latency",
"PGRecoveryContext queue latency");
osd_plb.add_u64(l_osd_loadavg, "loadavg", "CPU load");
osd_plb.add_u64(
l_osd_cached_crc, "cached_crc", "Total number getting crc from crc_cache");
osd_plb.add_u64(
l_osd_cached_crc_adjusted, "cached_crc_adjusted",
"Total number getting crc from crc_cache with adjusting");
osd_plb.add_u64(l_osd_missed_crc, "missed_crc",
"Total number of crc cache misses");
osd_plb.add_u64(l_osd_pg, "numpg", "Placement groups",
"pgs", PerfCountersBuilder::PRIO_USEFUL);
osd_plb.add_u64(
l_osd_pg_primary, "numpg_primary",
"Placement groups for which this osd is primary");
osd_plb.add_u64(
l_osd_pg_replica, "numpg_replica",
"Placement groups for which this osd is replica");
osd_plb.add_u64(
l_osd_pg_stray, "numpg_stray",
"Placement groups ready to be deleted from this osd");
osd_plb.add_u64(
l_osd_pg_removing, "numpg_removing",
"Placement groups queued for local deletion", "pgsr",
PerfCountersBuilder::PRIO_USEFUL);
osd_plb.add_u64(
l_osd_hb_to, "heartbeat_to_peers", "Heartbeat (ping) peers we send to");
osd_plb.add_u64_counter(l_osd_map, "map_messages", "OSD map messages");
osd_plb.add_u64_counter(l_osd_mape, "map_message_epochs", "OSD map epochs");
osd_plb.add_u64_counter(
l_osd_mape_dup, "map_message_epoch_dups", "OSD map duplicates");
osd_plb.add_u64_counter(
l_osd_waiting_for_map, "messages_delayed_for_map",
"Operations waiting for OSD map");
osd_plb.add_u64_counter(
l_osd_map_cache_hit, "osd_map_cache_hit", "osdmap cache hit");
osd_plb.add_u64_counter(
l_osd_map_cache_miss, "osd_map_cache_miss", "osdmap cache miss");
osd_plb.add_u64_counter(
l_osd_map_cache_miss_low, "osd_map_cache_miss_low",
"osdmap cache miss below cache lower bound");
osd_plb.add_u64_avg(
l_osd_map_cache_miss_low_avg, "osd_map_cache_miss_low_avg",
"osdmap cache miss, avg distance below cache lower bound");
osd_plb.add_u64_counter(
l_osd_map_bl_cache_hit, "osd_map_bl_cache_hit",
"OSDMap buffer cache hits");
osd_plb.add_u64_counter(
l_osd_map_bl_cache_miss, "osd_map_bl_cache_miss",
"OSDMap buffer cache misses");
osd_plb.add_u64(
l_osd_stat_bytes, "stat_bytes", "OSD size", "size",
PerfCountersBuilder::PRIO_USEFUL, unit_t(UNIT_BYTES));
osd_plb.add_u64(
l_osd_stat_bytes_used, "stat_bytes_used", "Used space", "used",
PerfCountersBuilder::PRIO_USEFUL, unit_t(UNIT_BYTES));
osd_plb.add_u64(l_osd_stat_bytes_avail, "stat_bytes_avail", "Available space", NULL, 0, unit_t(UNIT_BYTES));
osd_plb.add_u64_counter(
l_osd_copyfrom, "copyfrom", "Rados \"copy-from\" operations");
osd_plb.add_u64_counter(l_osd_tier_promote, "tier_promote", "Tier promotions");
osd_plb.add_u64_counter(l_osd_tier_flush, "tier_flush", "Tier flushes");
osd_plb.add_u64_counter(
l_osd_tier_flush_fail, "tier_flush_fail", "Failed tier flushes");
osd_plb.add_u64_counter(
l_osd_tier_try_flush, "tier_try_flush", "Tier flush attempts");
osd_plb.add_u64_counter(
l_osd_tier_try_flush_fail, "tier_try_flush_fail",
"Failed tier flush attempts");
osd_plb.add_u64_counter(
l_osd_tier_evict, "tier_evict", "Tier evictions");
osd_plb.add_u64_counter(
l_osd_tier_whiteout, "tier_whiteout", "Tier whiteouts");
osd_plb.add_u64_counter(
l_osd_tier_dirty, "tier_dirty", "Dirty tier flag set");
osd_plb.add_u64_counter(
l_osd_tier_clean, "tier_clean", "Dirty tier flag cleaned");
osd_plb.add_u64_counter(
l_osd_tier_delay, "tier_delay", "Tier delays (agent waiting)");
osd_plb.add_u64_counter(
l_osd_tier_proxy_read, "tier_proxy_read", "Tier proxy reads");
osd_plb.add_u64_counter(
l_osd_tier_proxy_write, "tier_proxy_write", "Tier proxy writes");
osd_plb.add_u64_counter(
l_osd_agent_wake, "agent_wake", "Tiering agent wake up");
osd_plb.add_u64_counter(
l_osd_agent_skip, "agent_skip", "Objects skipped by agent");
osd_plb.add_u64_counter(
l_osd_agent_flush, "agent_flush", "Tiering agent flushes");
osd_plb.add_u64_counter(
l_osd_agent_evict, "agent_evict", "Tiering agent evictions");
osd_plb.add_u64_counter(
l_osd_object_ctx_cache_hit, "object_ctx_cache_hit", "Object context cache hits");
osd_plb.add_u64_counter(
l_osd_object_ctx_cache_total, "object_ctx_cache_total", "Object context cache lookups");
osd_plb.add_u64_counter(l_osd_op_cache_hit, "op_cache_hit");
osd_plb.add_time_avg(
l_osd_tier_flush_lat, "osd_tier_flush_lat", "Object flush latency");
osd_plb.add_time_avg(
l_osd_tier_promote_lat, "osd_tier_promote_lat", "Object promote latency");
osd_plb.add_time_avg(
l_osd_tier_r_lat, "osd_tier_r_lat", "Object proxy read latency");
osd_plb.add_u64_counter(
l_osd_pg_info, "osd_pg_info", "PG updated its info (using any method)");
osd_plb.add_u64_counter(
l_osd_pg_fastinfo, "osd_pg_fastinfo",
"PG updated its info using fastinfo attr");
osd_plb.add_u64_counter(
l_osd_pg_biginfo, "osd_pg_biginfo", "PG updated its biginfo attr");
return osd_plb.create_perf_counters();
}
PerfCounters *build_recoverystate_perf(CephContext *cct) {
PerfCountersBuilder rs_perf(cct, "recoverystate_perf", rs_first, rs_last);
rs_perf.add_time_avg(rs_initial_latency, "initial_latency", "Initial recovery state latency");
rs_perf.add_time_avg(rs_started_latency, "started_latency", "Started recovery state latency");
rs_perf.add_time_avg(rs_reset_latency, "reset_latency", "Reset recovery state latency");
rs_perf.add_time_avg(rs_start_latency, "start_latency", "Start recovery state latency");
rs_perf.add_time_avg(rs_primary_latency, "primary_latency", "Primary recovery state latency");
rs_perf.add_time_avg(rs_peering_latency, "peering_latency", "Peering recovery state latency");
rs_perf.add_time_avg(rs_backfilling_latency, "backfilling_latency", "Backfilling recovery state latency");
rs_perf.add_time_avg(rs_waitremotebackfillreserved_latency, "waitremotebackfillreserved_latency", "Wait remote backfill reserved recovery state latency");
rs_perf.add_time_avg(rs_waitlocalbackfillreserved_latency, "waitlocalbackfillreserved_latency", "Wait local backfill reserved recovery state latency");
rs_perf.add_time_avg(rs_notbackfilling_latency, "notbackfilling_latency", "Notbackfilling recovery state latency");
rs_perf.add_time_avg(rs_repnotrecovering_latency, "repnotrecovering_latency", "Repnotrecovering recovery state latency");
rs_perf.add_time_avg(rs_repwaitrecoveryreserved_latency, "repwaitrecoveryreserved_latency", "Rep wait recovery reserved recovery state latency");
rs_perf.add_time_avg(rs_repwaitbackfillreserved_latency, "repwaitbackfillreserved_latency", "Rep wait backfill reserved recovery state latency");
rs_perf.add_time_avg(rs_reprecovering_latency, "reprecovering_latency", "RepRecovering recovery state latency");
rs_perf.add_time_avg(rs_activating_latency, "activating_latency", "Activating recovery state latency");
rs_perf.add_time_avg(rs_waitlocalrecoveryreserved_latency, "waitlocalrecoveryreserved_latency", "Wait local recovery reserved recovery state latency");
rs_perf.add_time_avg(rs_waitremoterecoveryreserved_latency, "waitremoterecoveryreserved_latency", "Wait remote recovery reserved recovery state latency");
rs_perf.add_time_avg(rs_recovering_latency, "recovering_latency", "Recovering recovery state latency");
rs_perf.add_time_avg(rs_recovered_latency, "recovered_latency", "Recovered recovery state latency");
rs_perf.add_time_avg(rs_clean_latency, "clean_latency", "Clean recovery state latency");
rs_perf.add_time_avg(rs_active_latency, "active_latency", "Active recovery state latency");
rs_perf.add_time_avg(rs_replicaactive_latency, "replicaactive_latency", "Replicaactive recovery state latency");
rs_perf.add_time_avg(rs_stray_latency, "stray_latency", "Stray recovery state latency");
rs_perf.add_time_avg(rs_getinfo_latency, "getinfo_latency", "Getinfo recovery state latency");
rs_perf.add_time_avg(rs_getlog_latency, "getlog_latency", "Getlog recovery state latency");
rs_perf.add_time_avg(rs_waitactingchange_latency, "waitactingchange_latency", "Waitactingchange recovery state latency");
rs_perf.add_time_avg(rs_incomplete_latency, "incomplete_latency", "Incomplete recovery state latency");
rs_perf.add_time_avg(rs_down_latency, "down_latency", "Down recovery state latency");
rs_perf.add_time_avg(rs_getmissing_latency, "getmissing_latency", "Getmissing recovery state latency");
rs_perf.add_time_avg(rs_waitupthru_latency, "waitupthru_latency", "Waitupthru recovery state latency");
rs_perf.add_time_avg(rs_notrecovering_latency, "notrecovering_latency", "Notrecovering recovery state latency");
return rs_perf.create_perf_counters();
}
|
