| 1 | /* |
| 2 | * librdkafka - The Apache Kafka C/C++ library |
| 3 | * |
| 4 | * Copyright (c) 2015 Magnus Edenhill |
| 5 | * All rights reserved. |
| 6 | * |
| 7 | * Redistribution and use in source and binary forms, with or without |
| 8 | * modification, are permitted provided that the following conditions are met: |
| 9 | * |
| 10 | * 1. Redistributions of source code must retain the above copyright notice, |
| 11 | * this list of conditions and the following disclaimer. |
| 12 | * 2. Redistributions in binary form must reproduce the above copyright notice, |
| 13 | * this list of conditions and the following disclaimer in the documentation |
| 14 | * and/or other materials provided with the distribution. |
| 15 | * |
| 16 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
| 17 | * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 18 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 19 | * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE |
| 20 | * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
| 21 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
| 22 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
| 23 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
| 24 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
| 25 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
| 26 | * POSSIBILITY OF SUCH DAMAGE. |
| 27 | */ |
| 28 | #ifndef _RDKAFKA_PARTITION_H_ |
| 29 | #define _RDKAFKA_PARTITION_H_ |
| 30 | |
| 31 | #include "rdkafka_topic.h" |
| 32 | #include "rdkafka_cgrp.h" |
| 33 | #include "rdkafka_broker.h" |
| 34 | |
| 35 | extern const char *rd_kafka_fetch_states[]; |
| 36 | |
| 37 | |
| 38 | /** |
| 39 | * @brief Offset statistics |
| 40 | */ |
| 41 | struct offset_stats { |
| 42 | int64_t fetch_offset; /**< Next offset to fetch */ |
| 43 | int64_t eof_offset; /**< Last offset we reported EOF for */ |
| 44 | int64_t hi_offset; /**< Current broker hi offset */ |
| 45 | }; |
| 46 | |
| 47 | /** |
| 48 | * @brief Reset offset_stats struct to default values |
| 49 | */ |
| 50 | static RD_UNUSED void rd_kafka_offset_stats_reset (struct offset_stats *offs) { |
| 51 | offs->fetch_offset = 0; |
| 52 | offs->eof_offset = RD_KAFKA_OFFSET_INVALID; |
| 53 | offs->hi_offset = RD_KAFKA_OFFSET_INVALID; |
| 54 | } |
| 55 | |
| 56 | |
| 57 | /** |
| 58 | * @brief Store information about a partition error for future use. |
| 59 | */ |
| 60 | struct rd_kafka_toppar_err { |
| 61 | rd_kafka_resp_err_t err; /**< Error code */ |
| 62 | int actions; /**< Request actions */ |
| 63 | rd_ts_t ts; /**< Timestamp */ |
| 64 | uint64_t base_msgid; /**< First msg msgid */ |
| 65 | int32_t base_seq; /**< Idempodent Producer: |
| 66 | * first msg sequence */ |
| 67 | int32_t last_seq; /**< Idempotent Producer: |
| 68 | * last msg sequence */ |
| 69 | }; |
| 70 | |
| 71 | /** |
| 72 | * Topic + Partition combination |
| 73 | */ |
| 74 | struct rd_kafka_toppar_s { /* rd_kafka_toppar_t */ |
| 75 | TAILQ_ENTRY(rd_kafka_toppar_s) rktp_rklink; /* rd_kafka_t link */ |
| 76 | TAILQ_ENTRY(rd_kafka_toppar_s) rktp_rkblink; /* rd_kafka_broker_t link*/ |
| 77 | CIRCLEQ_ENTRY(rd_kafka_toppar_s) rktp_activelink; /* rkb_active_toppars */ |
| 78 | TAILQ_ENTRY(rd_kafka_toppar_s) rktp_rktlink; /* rd_kafka_itopic_t link*/ |
| 79 | TAILQ_ENTRY(rd_kafka_toppar_s) rktp_cgrplink;/* rd_kafka_cgrp_t link */ |
| 80 | rd_kafka_itopic_t *rktp_rkt; |
| 81 | shptr_rd_kafka_itopic_t *rktp_s_rkt; /* shared pointer for rktp_rkt */ |
| 82 | int32_t rktp_partition; |
| 83 | //LOCK: toppar_lock() + topic_wrlock() |
| 84 | //LOCK: .. in partition_available() |
| 85 | int32_t rktp_leader_id; /**< Current leader broker id. |
| 86 | * This is updated directly |
| 87 | * from metadata. */ |
| 88 | rd_kafka_broker_t *rktp_leader; /**< Current leader broker |
| 89 | * This updated asynchronously |
| 90 | * by issuing JOIN op to |
| 91 | * broker thread, so be careful |
| 92 | * in using this since it |
| 93 | * may lag. */ |
| 94 | rd_kafka_broker_t *rktp_next_leader; /**< Next leader broker after |
| 95 | * async migration op. */ |
| 96 | rd_refcnt_t rktp_refcnt; |
| 97 | mtx_t rktp_lock; |
| 98 | |
| 99 | //LOCK: toppar_lock. toppar_insert_msg(), concat_msgq() |
| 100 | //LOCK: toppar_lock. toppar_enq_msg(), deq_msg(), toppar_retry_msgq() |
| 101 | rd_kafka_q_t *rktp_msgq_wakeup_q; /**< Wake-up queue */ |
| 102 | rd_kafka_msgq_t rktp_msgq; /* application->rdkafka queue. |
| 103 | * protected by rktp_lock */ |
| 104 | rd_kafka_msgq_t rktp_xmit_msgq; /* internal broker xmit queue. |
| 105 | * local to broker thread. */ |
| 106 | |
| 107 | int rktp_fetch; /* On rkb_active_toppars list */ |
| 108 | |
| 109 | /* Consumer */ |
| 110 | rd_kafka_q_t *rktp_fetchq; /* Queue of fetched messages |
| 111 | * from broker. |
| 112 | * Broker thread -> App */ |
| 113 | rd_kafka_q_t *rktp_ops; /* * -> Main thread */ |
| 114 | |
| 115 | rd_atomic32_t rktp_msgs_inflight; /**< Current number of |
| 116 | * messages in-flight to/from |
| 117 | * the broker. */ |
| 118 | |
| 119 | uint64_t rktp_msgid; /**< Current/last message id. |
| 120 | * Each message enqueued on a |
| 121 | * non-UA partition will get a |
| 122 | * partition-unique sequencial |
| 123 | * number assigned. |
| 124 | * This number is used to |
| 125 | * re-enqueue the message |
| 126 | * on resends but making sure |
| 127 | * the input ordering is still |
| 128 | * maintained, and used by |
| 129 | * the idempotent producer. |
| 130 | * Starts at 1. |
| 131 | * Protected by toppar_lock */ |
| 132 | struct { |
| 133 | rd_kafka_pid_t pid; /**< Partition's last known |
| 134 | * Producer Id and epoch. |
| 135 | * Protected by toppar lock. |
| 136 | * Only updated in toppar |
| 137 | * handler thread. */ |
| 138 | uint64_t acked_msgid; /**< Highest acknowledged message. |
| 139 | * Protected by toppar lock. */ |
| 140 | uint64_t epoch_base_msgid; /**< This Producer epoch's |
| 141 | * base msgid. |
| 142 | * When a new epoch is |
| 143 | * acquired the base_seq |
| 144 | * is set to the current |
| 145 | * rktp_msgid so that |
| 146 | * sub-sequent produce |
| 147 | * requests will have |
| 148 | * a sequence number series |
| 149 | * starting at 0. |
| 150 | * Only accessed from |
| 151 | * toppar handler thread. */ |
| 152 | int32_t next_ack_seq; /**< Next expected ack sequence. |
| 153 | * Protected by toppar lock. */ |
| 154 | int32_t next_err_seq; /**< Next expected error sequence. |
| 155 | * Used when draining outstanding |
| 156 | * issues. |
| 157 | * This value will be the same |
| 158 | * as next_ack_seq until a drainable |
| 159 | * error occurs, in which case it |
| 160 | * will advance past next_ack_seq. |
| 161 | * next_ack_seq can never be larger |
| 162 | * than next_err_seq. |
| 163 | * Protected by toppar lock. */ |
| 164 | rd_bool_t wait_drain; /**< All inflight requests must |
| 165 | * be drained/finish before |
| 166 | * resuming producing. |
| 167 | * This is set to true |
| 168 | * when a leader change |
| 169 | * happens so that the |
| 170 | * in-flight messages for the |
| 171 | * old brokers finish before |
| 172 | * the new broker starts sending. |
| 173 | * This as a step to ensure |
| 174 | * consistency. |
| 175 | * Only accessed from toppar |
| 176 | * handler thread. */ |
| 177 | } rktp_eos; |
| 178 | |
| 179 | /** |
| 180 | * rktp version barriers |
| 181 | * |
| 182 | * rktp_version is the application/controller side's |
| 183 | * authoritative version, it depicts the most up to date state. |
| 184 | * This is what q_filter() matches an rko_version to. |
| 185 | * |
| 186 | * rktp_op_version is the last/current received state handled |
| 187 | * by the toppar in the broker thread. It is updated to rktp_version |
| 188 | * when receiving a new op. |
| 189 | * |
| 190 | * rktp_fetch_version is the current fetcher decision version. |
| 191 | * It is used in fetch_decide() to see if the fetch decision |
| 192 | * needs to be updated by comparing to rktp_op_version. |
| 193 | * |
| 194 | * Example: |
| 195 | * App thread : Send OP_START (v1 bump): rktp_version=1 |
| 196 | * Broker thread: Recv OP_START (v1): rktp_op_version=1 |
| 197 | * Broker thread: fetch_decide() detects that |
| 198 | * rktp_op_version != rktp_fetch_version and |
| 199 | * sets rktp_fetch_version=1. |
| 200 | * Broker thread: next Fetch request has it's tver state set to |
| 201 | * rktp_fetch_verison (v1). |
| 202 | * |
| 203 | * App thread : Send OP_SEEK (v2 bump): rktp_version=2 |
| 204 | * Broker thread: Recv OP_SEEK (v2): rktp_op_version=2 |
| 205 | * Broker thread: Recv IO FetchResponse with tver=1, |
| 206 | * when enqueued on rktp_fetchq they're discarded |
| 207 | * due to old version (tver<rktp_version). |
| 208 | * Broker thread: fetch_decide() detects version change and |
| 209 | * sets rktp_fetch_version=2. |
| 210 | * Broker thread: next Fetch request has tver=2 |
| 211 | * Broker thread: Recv IO FetchResponse with tver=2 which |
| 212 | * is same as rktp_version so message is forwarded |
| 213 | * to app. |
| 214 | */ |
| 215 | rd_atomic32_t rktp_version; /* Latest op version. |
| 216 | * Authoritative (app thread)*/ |
| 217 | int32_t rktp_op_version; /* Op version of curr command |
| 218 | * state from. |
| 219 | * (broker thread) */ |
| 220 | int32_t rktp_fetch_version; /* Op version of curr fetch. |
| 221 | (broker thread) */ |
| 222 | |
| 223 | enum { |
| 224 | RD_KAFKA_TOPPAR_FETCH_NONE = 0, |
| 225 | RD_KAFKA_TOPPAR_FETCH_STOPPING, |
| 226 | RD_KAFKA_TOPPAR_FETCH_STOPPED, |
| 227 | RD_KAFKA_TOPPAR_FETCH_OFFSET_QUERY, |
| 228 | RD_KAFKA_TOPPAR_FETCH_OFFSET_WAIT, |
| 229 | RD_KAFKA_TOPPAR_FETCH_ACTIVE, |
| 230 | } rktp_fetch_state; /* Broker thread's state */ |
| 231 | |
| 232 | #define RD_KAFKA_TOPPAR_FETCH_IS_STARTED(fetch_state) \ |
| 233 | ((fetch_state) >= RD_KAFKA_TOPPAR_FETCH_OFFSET_QUERY) |
| 234 | |
| 235 | int32_t rktp_fetch_msg_max_bytes; /* Max number of bytes to |
| 236 | * fetch. |
| 237 | * Locality: broker thread |
| 238 | */ |
| 239 | |
| 240 | rd_ts_t rktp_ts_fetch_backoff; /* Back off fetcher for |
| 241 | * this partition until this |
| 242 | * absolute timestamp |
| 243 | * expires. */ |
| 244 | |
| 245 | int64_t rktp_query_offset; /* Offset to query broker for*/ |
| 246 | int64_t rktp_next_offset; /* Next offset to start |
| 247 | * fetching from. |
| 248 | * Locality: toppar thread */ |
| 249 | int64_t rktp_last_next_offset; /* Last next_offset handled |
| 250 | * by fetch_decide(). |
| 251 | * Locality: broker thread */ |
| 252 | int64_t rktp_app_offset; /* Last offset delivered to |
| 253 | * application + 1. |
| 254 | * Is reset to INVALID_OFFSET |
| 255 | * when partition is |
| 256 | * unassigned/stopped. */ |
| 257 | int64_t rktp_stored_offset; /* Last stored offset, but |
| 258 | * maybe not committed yet. */ |
| 259 | int64_t rktp_committing_offset; /* Offset currently being |
| 260 | * committed */ |
| 261 | int64_t rktp_committed_offset; /* Last committed offset */ |
| 262 | rd_ts_t rktp_ts_committed_offset; /* Timestamp of last |
| 263 | * commit */ |
| 264 | |
| 265 | struct offset_stats rktp_offsets; /* Current offsets. |
| 266 | * Locality: broker thread*/ |
| 267 | struct offset_stats rktp_offsets_fin; /* Finalized offset for stats. |
| 268 | * Updated periodically |
| 269 | * by broker thread. |
| 270 | * Locks: toppar_lock */ |
| 271 | |
| 272 | int64_t rktp_hi_offset; /* Current high offset. |
| 273 | * Locks: toppar_lock */ |
| 274 | int64_t rktp_lo_offset; /* Current broker low offset. |
| 275 | * This is outside of the stats |
| 276 | * struct due to this field |
| 277 | * being populated by the |
| 278 | * toppar thread rather than |
| 279 | * the broker thread. |
| 280 | * Locality: toppar thread |
| 281 | * Locks: toppar_lock */ |
| 282 | |
| 283 | rd_ts_t rktp_ts_offset_lag; |
| 284 | |
| 285 | char *rktp_offset_path; /* Path to offset file */ |
| 286 | FILE *rktp_offset_fp; /* Offset file pointer */ |
| 287 | rd_kafka_cgrp_t *rktp_cgrp; /* Belongs to this cgrp */ |
| 288 | |
| 289 | int rktp_assigned; /* Partition in cgrp assignment */ |
| 290 | |
| 291 | rd_kafka_replyq_t rktp_replyq; /* Current replyq+version |
| 292 | * for propagating |
| 293 | * major operations, e.g., |
| 294 | * FETCH_STOP. */ |
| 295 | //LOCK: toppar_lock(). RD_KAFKA_TOPPAR_F_DESIRED |
| 296 | //LOCK: toppar_lock(). RD_KAFKA_TOPPAR_F_UNKNOWN |
| 297 | int rktp_flags; |
| 298 | #define RD_KAFKA_TOPPAR_F_DESIRED 0x1 /* This partition is desired |
| 299 | * by a consumer. */ |
| 300 | #define RD_KAFKA_TOPPAR_F_UNKNOWN 0x2 /* Topic is not yet or no longer |
| 301 | * seen on a broker. */ |
| 302 | #define RD_KAFKA_TOPPAR_F_OFFSET_STORE 0x4 /* Offset store is active */ |
| 303 | #define RD_KAFKA_TOPPAR_F_OFFSET_STORE_STOPPING 0x8 /* Offset store stopping */ |
| 304 | #define RD_KAFKA_TOPPAR_F_APP_PAUSE 0x10 /* App pause()d consumption */ |
| 305 | #define RD_KAFKA_TOPPAR_F_LIB_PAUSE 0x20 /* librdkafka paused consumption */ |
| 306 | #define RD_KAFKA_TOPPAR_F_REMOVE 0x40 /* partition removed from cluster */ |
| 307 | #define RD_KAFKA_TOPPAR_F_LEADER_ERR 0x80 /* Operation failed: |
| 308 | * leader might be missing. |
| 309 | * Typically set from |
| 310 | * ProduceResponse failure. */ |
| 311 | |
| 312 | shptr_rd_kafka_toppar_t *rktp_s_for_desp; /* Shared pointer for |
| 313 | * rkt_desp list */ |
| 314 | shptr_rd_kafka_toppar_t *rktp_s_for_cgrp; /* Shared pointer for |
| 315 | * rkcg_toppars list */ |
| 316 | shptr_rd_kafka_toppar_t *rktp_s_for_rkb; /* Shared pointer for |
| 317 | * rkb_toppars list */ |
| 318 | |
| 319 | /* |
| 320 | * Timers |
| 321 | */ |
| 322 | rd_kafka_timer_t rktp_offset_query_tmr; /* Offset query timer */ |
| 323 | rd_kafka_timer_t rktp_offset_commit_tmr; /* Offset commit timer */ |
| 324 | rd_kafka_timer_t rktp_offset_sync_tmr; /* Offset file sync timer */ |
| 325 | rd_kafka_timer_t rktp_consumer_lag_tmr; /* Consumer lag monitoring |
| 326 | * timer */ |
| 327 | |
| 328 | int rktp_wait_consumer_lag_resp; /* Waiting for consumer lag |
| 329 | * response. */ |
| 330 | |
| 331 | struct rd_kafka_toppar_err rktp_last_err; /**< Last produce error */ |
| 332 | |
| 333 | |
| 334 | struct { |
| 335 | rd_atomic64_t tx_msgs; /**< Producer: sent messages */ |
| 336 | rd_atomic64_t tx_msg_bytes; /**< .. bytes */ |
| 337 | rd_atomic64_t rx_msgs; /**< Consumer: received messages */ |
| 338 | rd_atomic64_t rx_msg_bytes; /**< .. bytes */ |
| 339 | rd_atomic64_t producer_enq_msgs; /**< Producer: enqueued msgs */ |
| 340 | rd_atomic64_t rx_ver_drops; /**< Consumer: outdated message |
| 341 | * drops. */ |
| 342 | } rktp_c; |
| 343 | |
| 344 | }; |
| 345 | |
| 346 | |
| 347 | /** |
| 348 | * Check if toppar is paused (consumer). |
| 349 | * Locks: toppar_lock() MUST be held. |
| 350 | */ |
| 351 | #define RD_KAFKA_TOPPAR_IS_PAUSED(rktp) \ |
| 352 | ((rktp)->rktp_flags & (RD_KAFKA_TOPPAR_F_APP_PAUSE | \ |
| 353 | RD_KAFKA_TOPPAR_F_LIB_PAUSE)) |
| 354 | |
| 355 | |
| 356 | |
| 357 | |
| 358 | /* Converts a shptr..toppar_t to a toppar_t */ |
| 359 | #define rd_kafka_toppar_s2i(s_rktp) rd_shared_ptr_obj(s_rktp) |
| 360 | |
| 361 | |
| 362 | /** |
| 363 | * Returns a shared pointer for the topic. |
| 364 | */ |
| 365 | #define rd_kafka_toppar_keep(rktp) \ |
| 366 | rd_shared_ptr_get(rktp, &(rktp)->rktp_refcnt, shptr_rd_kafka_toppar_t) |
| 367 | |
| 368 | #define rd_kafka_toppar_keep_src(func,line,rktp) \ |
| 369 | rd_shared_ptr_get_src(func, line, rktp, \ |
| 370 | &(rktp)->rktp_refcnt, shptr_rd_kafka_toppar_t) |
| 371 | |
| 372 | |
| 373 | /** |
| 374 | * Frees a shared pointer previously returned by ..toppar_keep() |
| 375 | */ |
| 376 | #define rd_kafka_toppar_destroy(s_rktp) \ |
| 377 | rd_shared_ptr_put(s_rktp, \ |
| 378 | &rd_kafka_toppar_s2i(s_rktp)->rktp_refcnt, \ |
| 379 | rd_kafka_toppar_destroy_final( \ |
| 380 | rd_kafka_toppar_s2i(s_rktp))) |
| 381 | |
| 382 | |
| 383 | |
| 384 | |
| 385 | #define rd_kafka_toppar_lock(rktp) mtx_lock(&(rktp)->rktp_lock) |
| 386 | #define rd_kafka_toppar_unlock(rktp) mtx_unlock(&(rktp)->rktp_lock) |
| 387 | |
| 388 | static const char *rd_kafka_toppar_name (const rd_kafka_toppar_t *rktp) |
| 389 | RD_UNUSED; |
| 390 | static const char *rd_kafka_toppar_name (const rd_kafka_toppar_t *rktp) { |
| 391 | static RD_TLS char ret[256]; |
| 392 | |
| 393 | rd_snprintf(ret, sizeof(ret), "%.*s [%" PRId32"]" , |
| 394 | RD_KAFKAP_STR_PR(rktp->rktp_rkt->rkt_topic), |
| 395 | rktp->rktp_partition); |
| 396 | |
| 397 | return ret; |
| 398 | } |
| 399 | shptr_rd_kafka_toppar_t *rd_kafka_toppar_new0 (rd_kafka_itopic_t *rkt, |
| 400 | int32_t partition, |
| 401 | const char *func, int line); |
| 402 | #define rd_kafka_toppar_new(rkt,partition) \ |
| 403 | rd_kafka_toppar_new0(rkt, partition, __FUNCTION__, __LINE__) |
| 404 | void rd_kafka_toppar_destroy_final (rd_kafka_toppar_t *rktp); |
| 405 | void rd_kafka_toppar_purge_queues (rd_kafka_toppar_t *rktp); |
| 406 | void rd_kafka_toppar_set_fetch_state (rd_kafka_toppar_t *rktp, |
| 407 | int fetch_state); |
| 408 | void rd_kafka_toppar_insert_msg (rd_kafka_toppar_t *rktp, rd_kafka_msg_t *rkm); |
| 409 | void rd_kafka_toppar_enq_msg (rd_kafka_toppar_t *rktp, rd_kafka_msg_t *rkm); |
| 410 | int rd_kafka_retry_msgq (rd_kafka_msgq_t *destq, |
| 411 | rd_kafka_msgq_t *srcq, |
| 412 | int incr_retry, int max_retries, rd_ts_t backoff, |
| 413 | rd_kafka_msg_status_t status, |
| 414 | int (*cmp) (const void *a, const void *b)); |
| 415 | void rd_kafka_msgq_insert_msgq (rd_kafka_msgq_t *destq, |
| 416 | rd_kafka_msgq_t *srcq, |
| 417 | int (*cmp) (const void *a, const void *b)); |
| 418 | int rd_kafka_toppar_retry_msgq (rd_kafka_toppar_t *rktp, |
| 419 | rd_kafka_msgq_t *rkmq, |
| 420 | int incr_retry, rd_kafka_msg_status_t status); |
| 421 | void rd_kafka_toppar_insert_msgq (rd_kafka_toppar_t *rktp, |
| 422 | rd_kafka_msgq_t *rkmq); |
| 423 | void rd_kafka_toppar_enq_error (rd_kafka_toppar_t *rktp, |
| 424 | rd_kafka_resp_err_t err, |
| 425 | const char *reason); |
| 426 | shptr_rd_kafka_toppar_t *rd_kafka_toppar_get0 (const char *func, int line, |
| 427 | const rd_kafka_itopic_t *rkt, |
| 428 | int32_t partition, |
| 429 | int ua_on_miss); |
| 430 | #define rd_kafka_toppar_get(rkt,partition,ua_on_miss) \ |
| 431 | rd_kafka_toppar_get0(__FUNCTION__,__LINE__,rkt,partition,ua_on_miss) |
| 432 | shptr_rd_kafka_toppar_t *rd_kafka_toppar_get2 (rd_kafka_t *rk, |
| 433 | const char *topic, |
| 434 | int32_t partition, |
| 435 | int ua_on_miss, |
| 436 | int create_on_miss); |
| 437 | shptr_rd_kafka_toppar_t * |
| 438 | rd_kafka_toppar_get_avail (const rd_kafka_itopic_t *rkt, |
| 439 | int32_t partition, |
| 440 | int ua_on_miss, |
| 441 | rd_kafka_resp_err_t *errp); |
| 442 | |
| 443 | shptr_rd_kafka_toppar_t *rd_kafka_toppar_desired_get (rd_kafka_itopic_t *rkt, |
| 444 | int32_t partition); |
| 445 | void rd_kafka_toppar_desired_add0 (rd_kafka_toppar_t *rktp); |
| 446 | shptr_rd_kafka_toppar_t *rd_kafka_toppar_desired_add (rd_kafka_itopic_t *rkt, |
| 447 | int32_t partition); |
| 448 | void rd_kafka_toppar_desired_link (rd_kafka_toppar_t *rktp); |
| 449 | void rd_kafka_toppar_desired_unlink (rd_kafka_toppar_t *rktp); |
| 450 | void rd_kafka_toppar_desired_del (rd_kafka_toppar_t *rktp); |
| 451 | |
| 452 | void rd_kafka_toppar_next_offset_handle (rd_kafka_toppar_t *rktp, |
| 453 | int64_t Offset); |
| 454 | |
| 455 | void rd_kafka_toppar_offset_commit (rd_kafka_toppar_t *rktp, int64_t offset, |
| 456 | const char *metadata); |
| 457 | |
| 458 | void rd_kafka_toppar_broker_delegate (rd_kafka_toppar_t *rktp, |
| 459 | rd_kafka_broker_t *rkb, |
| 460 | int for_removal); |
| 461 | |
| 462 | |
| 463 | rd_kafka_resp_err_t rd_kafka_toppar_op_fetch_start (rd_kafka_toppar_t *rktp, |
| 464 | int64_t offset, |
| 465 | rd_kafka_q_t *fwdq, |
| 466 | rd_kafka_replyq_t replyq); |
| 467 | |
| 468 | rd_kafka_resp_err_t rd_kafka_toppar_op_fetch_stop (rd_kafka_toppar_t *rktp, |
| 469 | rd_kafka_replyq_t replyq); |
| 470 | |
| 471 | rd_kafka_resp_err_t rd_kafka_toppar_op_seek (rd_kafka_toppar_t *rktp, |
| 472 | int64_t offset, |
| 473 | rd_kafka_replyq_t replyq); |
| 474 | |
| 475 | rd_kafka_resp_err_t rd_kafka_toppar_op_pause (rd_kafka_toppar_t *rktp, |
| 476 | int pause, int flag); |
| 477 | |
| 478 | void rd_kafka_toppar_fetch_stopped (rd_kafka_toppar_t *rktp, |
| 479 | rd_kafka_resp_err_t err); |
| 480 | |
| 481 | |
| 482 | |
| 483 | rd_ts_t rd_kafka_toppar_fetch_decide (rd_kafka_toppar_t *rktp, |
| 484 | rd_kafka_broker_t *rkb, |
| 485 | int force_remove); |
| 486 | |
| 487 | |
| 488 | |
| 489 | rd_ts_t rd_kafka_broker_consumer_toppar_serve (rd_kafka_broker_t *rkb, |
| 490 | rd_kafka_toppar_t *rktp); |
| 491 | |
| 492 | |
| 493 | void rd_kafka_toppar_offset_fetch (rd_kafka_toppar_t *rktp, |
| 494 | rd_kafka_replyq_t replyq); |
| 495 | |
| 496 | void rd_kafka_toppar_offset_request (rd_kafka_toppar_t *rktp, |
| 497 | int64_t query_offset, int backoff_ms); |
| 498 | |
| 499 | |
| 500 | rd_kafka_assignor_t * |
| 501 | rd_kafka_assignor_find (rd_kafka_t *rk, const char *protocol); |
| 502 | |
| 503 | |
| 504 | rd_kafka_broker_t *rd_kafka_toppar_leader (rd_kafka_toppar_t *rktp, |
| 505 | int proper_broker); |
| 506 | void rd_kafka_toppar_leader_unavailable (rd_kafka_toppar_t *rktp, |
| 507 | const char *reason, |
| 508 | rd_kafka_resp_err_t err); |
| 509 | |
| 510 | rd_kafka_resp_err_t |
| 511 | rd_kafka_toppars_pause_resume (rd_kafka_t *rk, int pause, int flag, |
| 512 | rd_kafka_topic_partition_list_t *partitions); |
| 513 | |
| 514 | |
| 515 | rd_kafka_topic_partition_t *rd_kafka_topic_partition_new (const char *topic, |
| 516 | int32_t partition); |
| 517 | rd_kafka_topic_partition_t * |
| 518 | rd_kafka_topic_partition_new_from_rktp (rd_kafka_toppar_t *rktp); |
| 519 | |
| 520 | rd_kafka_topic_partition_t * |
| 521 | rd_kafka_topic_partition_list_add0 (rd_kafka_topic_partition_list_t *rktparlist, |
| 522 | const char *topic, int32_t partition, |
| 523 | shptr_rd_kafka_toppar_t *_private); |
| 524 | |
| 525 | rd_kafka_topic_partition_t * |
| 526 | rd_kafka_topic_partition_list_upsert ( |
| 527 | rd_kafka_topic_partition_list_t *rktparlist, |
| 528 | const char *topic, int32_t partition); |
| 529 | |
| 530 | int rd_kafka_topic_partition_match (rd_kafka_t *rk, |
| 531 | const rd_kafka_group_member_t *rkgm, |
| 532 | const rd_kafka_topic_partition_t *rktpar, |
| 533 | const char *topic, int *matched_by_regex); |
| 534 | |
| 535 | |
| 536 | void rd_kafka_topic_partition_list_sort_by_topic ( |
| 537 | rd_kafka_topic_partition_list_t *rktparlist); |
| 538 | |
| 539 | void |
| 540 | rd_kafka_topic_partition_list_reset_offsets (rd_kafka_topic_partition_list_t *rktparlist, |
| 541 | int64_t offset); |
| 542 | |
| 543 | int rd_kafka_topic_partition_list_set_offsets ( |
| 544 | rd_kafka_t *rk, |
| 545 | rd_kafka_topic_partition_list_t *rktparlist, |
| 546 | int from_rktp, int64_t def_value, int is_commit); |
| 547 | |
| 548 | int rd_kafka_topic_partition_list_count_abs_offsets ( |
| 549 | const rd_kafka_topic_partition_list_t *rktparlist); |
| 550 | |
| 551 | shptr_rd_kafka_toppar_t * |
| 552 | rd_kafka_topic_partition_get_toppar (rd_kafka_t *rk, |
| 553 | rd_kafka_topic_partition_t *rktpar); |
| 554 | |
| 555 | shptr_rd_kafka_toppar_t * |
| 556 | rd_kafka_topic_partition_list_get_toppar ( |
| 557 | rd_kafka_t *rk, rd_kafka_topic_partition_t *rktpar); |
| 558 | |
| 559 | void |
| 560 | rd_kafka_topic_partition_list_update_toppars (rd_kafka_t *rk, |
| 561 | rd_kafka_topic_partition_list_t |
| 562 | *rktparlist); |
| 563 | |
| 564 | int |
| 565 | rd_kafka_topic_partition_list_get_leaders ( |
| 566 | rd_kafka_t *rk, |
| 567 | rd_kafka_topic_partition_list_t *rktparlist, |
| 568 | rd_list_t *leaders, rd_list_t *query_topics); |
| 569 | |
| 570 | rd_kafka_resp_err_t |
| 571 | rd_kafka_topic_partition_list_query_leaders ( |
| 572 | rd_kafka_t *rk, |
| 573 | rd_kafka_topic_partition_list_t *rktparlist, |
| 574 | rd_list_t *leaders, int timeout_ms); |
| 575 | |
| 576 | int |
| 577 | rd_kafka_topic_partition_list_get_topics ( |
| 578 | rd_kafka_t *rk, |
| 579 | rd_kafka_topic_partition_list_t *rktparlist, |
| 580 | rd_list_t *rkts); |
| 581 | |
| 582 | int |
| 583 | rd_kafka_topic_partition_list_get_topic_names ( |
| 584 | const rd_kafka_topic_partition_list_t *rktparlist, |
| 585 | rd_list_t *topics, int include_regex); |
| 586 | |
| 587 | void |
| 588 | rd_kafka_topic_partition_list_log (rd_kafka_t *rk, const char *fac, int dbg, |
| 589 | const rd_kafka_topic_partition_list_t *rktparlist); |
| 590 | |
| 591 | #define RD_KAFKA_FMT_F_OFFSET 0x1 /* Print offset */ |
| 592 | #define RD_KAFKA_FMT_F_ONLY_ERR 0x2 /* Only include errored entries */ |
| 593 | #define RD_KAFKA_FMT_F_NO_ERR 0x4 /* Dont print error string */ |
| 594 | const char * |
| 595 | rd_kafka_topic_partition_list_str (const rd_kafka_topic_partition_list_t *rktparlist, |
| 596 | char *dest, size_t dest_size, |
| 597 | int fmt_flags); |
| 598 | |
| 599 | void |
| 600 | rd_kafka_topic_partition_list_update (rd_kafka_topic_partition_list_t *dst, |
| 601 | const rd_kafka_topic_partition_list_t *src); |
| 602 | |
| 603 | int rd_kafka_topic_partition_leader_cmp (const void *_a, const void *_b); |
| 604 | |
| 605 | rd_kafka_topic_partition_list_t *rd_kafka_topic_partition_list_match ( |
| 606 | const rd_kafka_topic_partition_list_t *rktparlist, |
| 607 | int (*match) (const void *elem, const void *opaque), |
| 608 | void *opaque); |
| 609 | |
| 610 | size_t |
| 611 | rd_kafka_topic_partition_list_sum ( |
| 612 | const rd_kafka_topic_partition_list_t *rktparlist, |
| 613 | size_t (*cb) (const rd_kafka_topic_partition_t *rktpar, void *opaque), |
| 614 | void *opaque); |
| 615 | |
| 616 | void rd_kafka_topic_partition_list_set_err ( |
| 617 | rd_kafka_topic_partition_list_t *rktparlist, |
| 618 | rd_kafka_resp_err_t err); |
| 619 | |
| 620 | int rd_kafka_topic_partition_list_regex_cnt ( |
| 621 | const rd_kafka_topic_partition_list_t *rktparlist); |
| 622 | |
| 623 | /** |
| 624 | * @brief Toppar + Op version tuple used for mapping Fetched partitions |
| 625 | * back to their fetch versions. |
| 626 | */ |
| 627 | struct rd_kafka_toppar_ver { |
| 628 | shptr_rd_kafka_toppar_t *s_rktp; |
| 629 | int32_t version; |
| 630 | }; |
| 631 | |
| 632 | |
| 633 | /** |
| 634 | * @brief Toppar + Op version comparator. |
| 635 | */ |
| 636 | static RD_INLINE RD_UNUSED |
| 637 | int rd_kafka_toppar_ver_cmp (const void *_a, const void *_b) { |
| 638 | const struct rd_kafka_toppar_ver *a = _a, *b = _b; |
| 639 | const rd_kafka_toppar_t *rktp_a = rd_kafka_toppar_s2i(a->s_rktp); |
| 640 | const rd_kafka_toppar_t *rktp_b = rd_kafka_toppar_s2i(b->s_rktp); |
| 641 | int r; |
| 642 | |
| 643 | if (rktp_a->rktp_rkt != rktp_b->rktp_rkt && |
| 644 | (r = rd_kafkap_str_cmp(rktp_a->rktp_rkt->rkt_topic, |
| 645 | rktp_b->rktp_rkt->rkt_topic))) |
| 646 | return r; |
| 647 | |
| 648 | return rktp_a->rktp_partition - rktp_b->rktp_partition; |
| 649 | } |
| 650 | |
| 651 | /** |
| 652 | * @brief Frees up resources for \p tver but not the \p tver itself. |
| 653 | */ |
| 654 | static RD_INLINE RD_UNUSED |
| 655 | void rd_kafka_toppar_ver_destroy (struct rd_kafka_toppar_ver *tver) { |
| 656 | rd_kafka_toppar_destroy(tver->s_rktp); |
| 657 | } |
| 658 | |
| 659 | |
| 660 | /** |
| 661 | * @returns 1 if rko version is outdated, else 0. |
| 662 | */ |
| 663 | static RD_INLINE RD_UNUSED |
| 664 | int rd_kafka_op_version_outdated (rd_kafka_op_t *rko, int version) { |
| 665 | if (!rko->rko_version) |
| 666 | return 0; |
| 667 | |
| 668 | if (version) |
| 669 | return rko->rko_version < version; |
| 670 | |
| 671 | if (rko->rko_rktp) |
| 672 | return rko->rko_version < |
| 673 | rd_atomic32_get(&rd_kafka_toppar_s2i( |
| 674 | rko->rko_rktp)->rktp_version); |
| 675 | return 0; |
| 676 | } |
| 677 | |
| 678 | void |
| 679 | rd_kafka_toppar_offset_commit_result (rd_kafka_toppar_t *rktp, |
| 680 | rd_kafka_resp_err_t err, |
| 681 | rd_kafka_topic_partition_list_t *offsets); |
| 682 | |
| 683 | void rd_kafka_toppar_broker_leave_for_remove (rd_kafka_toppar_t *rktp); |
| 684 | |
| 685 | |
| 686 | /** |
| 687 | * @brief Represents a leader and the partitions it is leader for. |
| 688 | */ |
| 689 | struct rd_kafka_partition_leader { |
| 690 | rd_kafka_broker_t *rkb; |
| 691 | rd_kafka_topic_partition_list_t *partitions; |
| 692 | }; |
| 693 | |
| 694 | static RD_UNUSED void |
| 695 | rd_kafka_partition_leader_destroy (struct rd_kafka_partition_leader *leader) { |
| 696 | rd_kafka_broker_destroy(leader->rkb); |
| 697 | rd_kafka_topic_partition_list_destroy(leader->partitions); |
| 698 | rd_free(leader); |
| 699 | } |
| 700 | |
| 701 | static RD_UNUSED struct rd_kafka_partition_leader * |
| 702 | rd_kafka_partition_leader_new (rd_kafka_broker_t *rkb) { |
| 703 | struct rd_kafka_partition_leader *leader = rd_malloc(sizeof(*leader)); |
| 704 | leader->rkb = rkb; |
| 705 | rd_kafka_broker_keep(rkb); |
| 706 | leader->partitions = rd_kafka_topic_partition_list_new(0); |
| 707 | return leader; |
| 708 | } |
| 709 | |
| 710 | static RD_UNUSED |
| 711 | int rd_kafka_partition_leader_cmp (const void *_a, const void *_b) { |
| 712 | const struct rd_kafka_partition_leader *a = _a, *b = _b; |
| 713 | return rd_kafka_broker_cmp(a->rkb, b->rkb); |
| 714 | } |
| 715 | |
| 716 | int rd_kafka_toppar_pid_change (rd_kafka_toppar_t *rktp, rd_kafka_pid_t pid, |
| 717 | uint64_t base_msgid); |
| 718 | |
| 719 | int rd_kafka_toppar_handle_purge_queues (rd_kafka_toppar_t *rktp, |
| 720 | rd_kafka_broker_t *rkb, |
| 721 | int purge_flags); |
| 722 | void rd_kafka_purge_ua_toppar_queues (rd_kafka_t *rk); |
| 723 | |
| 724 | #endif /* _RDKAFKA_PARTITION_H_ */ |
| 725 | |