From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from [87.239.111.99] (localhost [127.0.0.1]) by dev.tarantool.org (Postfix) with ESMTP id 4B3AB6C7D2; Thu, 11 Feb 2021 15:17:57 +0300 (MSK) DKIM-Filter: OpenDKIM Filter v2.11.0 dev.tarantool.org 4B3AB6C7D2 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=tarantool.org; s=dev; t=1613045877; bh=sKtX8q+yM221NFr5DEqldp9AeB7eouZG86OQXVyXF8g=; h=To:Date:Subject:List-Id:List-Unsubscribe:List-Archive:List-Post: List-Help:List-Subscribe:From:Reply-To:Cc:From; b=mwWjzycX34c3Q5sekJkKb8xelPxBFfkmWg/7Oyw4OkBaezg2tSGpq/OfpLSoa8INg ++ha1CahFSNQm84fxgnYyq7mA0rPAyVtuQK/818jrB/1f2d9UeNKx/Ar0xysXcaFFV HTvp/y7coTth3clNmCoKMK/IEjRyF+Z2nI8dpV3k= Received: from smtp17.mail.ru (smtp17.mail.ru [94.100.176.154]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by dev.tarantool.org (Postfix) with ESMTPS id 5C8796C7D2 for ; Thu, 11 Feb 2021 15:17:56 +0300 (MSK) DKIM-Filter: OpenDKIM Filter v2.11.0 dev.tarantool.org 5C8796C7D2 Received: by smtp17.mail.ru with esmtpa (envelope-from ) id 1lAAuw-0000s8-D7; Thu, 11 Feb 2021 15:17:55 +0300 To: v.shpilevoy@tarantool.org, gorcunov@gmail.com Date: Thu, 11 Feb 2021 15:17:50 +0300 Message-Id: <20210211121750.46298-1-sergepetrenko@tarantool.org> X-Mailer: git-send-email 2.24.3 (Apple Git-128) MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-7564579A: B8F34718100C35BD X-77F55803: 4F1203BC0FB41BD9837AFC1E39A347B36CEC7F61DAD20D5BAEC12DBB1C9500CC182A05F5380850405B9CB0C35A124C3A4FBB954B2E4DAB835A947998A4FB459500CAEF412F110535 X-7FA49CB5: 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 X-C1DE0DAB: C20DE7B7AB408E4181F030C43753B8183A4AFAF3EA6BDC44E1F4276B80994196D27AC6BE5C1CDEA9CA15F8BDC6827D5F7BDCBACD4AA6F0EB9C2B6934AE262D3EE7EAB7254005DCED70F3E7AC9318472D1E0A4E2319210D9B64D260DF9561598F01A9E91200F654B06CE7B4E551862B828E8E86DC7131B365E7726E8460B7C23C X-C8649E89: 4E36BF7865823D7055A7F0CF078B5EC49A30900B95165D346C409ABC5F9C579B4F5B5F1E706EE8BB9ACF7889E7B2421CBD95F8C6AE75826724F3D6823F2381F11D7E09C32AA3244C60F391CFDB910D9DB68FE3F48ABAA515F94338140B71B8EE927AC6DF5659F194 X-D57D3AED: 3ZO7eAau8CL7WIMRKs4sN3D3tLDjz0dLbV79QFUyzQ2Ujvy7cMT6pYYqY16iZVKkSc3dCLJ7zSJH7+u4VD18S7Vl4ZUrpaVfd2+vE6kuoey4m4VkSEu530nj6fImhcD4MUrOEAnl0W826KZ9Q+tr5ycPtXkTV4k65bRjmOUUP8cvGozZ33TWg5HZplvhhXbhDGzqmQDTd6OAevLeAnq3Ra9uf7zvY2zzsIhlcp/Y7m53TZgf2aB4JOg4gkr2biojbL9S8ysBdXh1hhp/0dEluDzlSWaTegpJ X-Mailru-Sender: 3B9A0136629DC9125D61937A2360A446FD56A992E60266D5FE0F82D1A66D0772EC9509A64A923933424AE0EB1F3D1D21E2978F233C3FAE6EE63DB1732555E4A8EE80603BA4A5B0BC112434F685709FCF0DA7A0AF5A3A8387 X-Mras: Ok Subject: [Tarantool-patches] [PATCH v2] wal: introduce limits on simultaneous writes X-BeenThere: tarantool-patches@dev.tarantool.org X-Mailman-Version: 2.1.34 Precedence: list List-Id: Tarantool development patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , From: Serge Petrenko via Tarantool-patches Reply-To: Serge Petrenko Cc: tarantool-patches@dev.tarantool.org Errors-To: tarantool-patches-bounces@dev.tarantool.org Sender: "Tarantool-patches" Since the introduction of asynchronous commit, which doesn't wait for a WAL write to succeed, it's quite easy to clog WAL with huge amounts write requests. For now, it's only possible from an applier, since it's the only user of async commit at the moment. This happens when replica is syncing with master and reads new transactions at a pace higher than it can write them to WAL (see docbot request for detailed explanation). To ameliorate such behavior, we need to introduce some limit on not-yet-finished WAL write requests. This is what this commit is trying to do. Two new counters are added to wal writer: queue_size (in bytes) and queue_len (in wal messages) together with configuration settings: `wal_queue_max_size` and `wal_queue_max_len`. Size and length are increased on every new submitted request, and are decreased once the tx receives a confirmation that a specific request was written. Actually, the limits are added to an abstract journal, but take effect only for wal writer. Once size or length reach their maximum values, applier is blocked until some of the write requests are finished. The size limit isn't strict, i.e. if there's at least one free byte, the whole write request fits and no blocking is involved. The feature is ready for `box.commit{is_async=true}`. Once it's implemented, it should check whether the queue is full and let the user decide what to do next. Either wait or roll the tx back. Part of #5536 @TarantoolBot document Title: new configuration options: 'wal_queue_max_size', 'wal_queue_max_len' `wal_queue_max_size` and `wal_queue_max_len` put a limit on the amount of concurrent write requests submitted to WAL. `wal_queue_max_size` is measured in number of bytes to be written (0 means unlimited), and `wal_queue_max_len` is measured in number of transactions, 0 meaning unlimited. These options only affect replica behaviour at the moment, and default to 0. They limit the pace at which replica reads new transactions from master. Here's when these options come in handy: Imagine such a situation: there are 2 servers, a master and a replica, and the replica is down for some period of time. While the replica is down, the master serves requests at a reasonable pace, possibly close to its WAL throughput limit. Once the replica reconnects, it has to receive all the data master has piled up. Now there's no limit in speed at which master sends the data to replica, and there's no limit at which replica's applier submits corresponding write requests to WAL. This leads to a situation when replica's WAL is never in time to serve the requests and the amount of pending requests is constantly growing. There's no limit for memory WAL write requests take, and this clogging of WAL write queue may even lead to replica using up all the available memory. Now, when either `wal_queue_max_size` or `wal_queue_max_len` or both are set, appliers will stop reading new transactions once the limit is reached. This will let WAL process all the requests that have piled up and free all the excess memory. --- https://github.com/tarantool/tarantool/issues/5536 https://github.com/tarantool/tarantool/tree/sp/gh-5536-replica-oom Changes in v2: - Move queue logic to journal. src/box/applier.cc | 9 ++ src/box/box.cc | 46 +++++++ src/box/box.h | 2 + src/box/journal.c | 53 ++++++++ src/box/journal.h | 103 ++++++++++++++- src/box/lua/cfg.cc | 18 +++ src/box/lua/load_cfg.lua | 6 + src/box/wal.c | 18 +++ src/box/wal.h | 14 +++ test/app-tap/init_script.result | 2 + test/box-tap/cfg.test.lua | 4 +- test/box/admin.result | 4 + test/box/cfg.result | 8 ++ test/replication/gh-5536-wal-limit.result | 132 ++++++++++++++++++++ test/replication/gh-5536-wal-limit.test.lua | 58 +++++++++ test/replication/suite.cfg | 1 + test/replication/suite.ini | 2 +- 17 files changed, 472 insertions(+), 8 deletions(-) create mode 100644 test/replication/gh-5536-wal-limit.result create mode 100644 test/replication/gh-5536-wal-limit.test.lua diff --git a/src/box/applier.cc b/src/box/applier.cc index 553db76fc..06aaa0a79 100644 --- a/src/box/applier.cc +++ b/src/box/applier.cc @@ -967,6 +967,15 @@ applier_apply_tx(struct applier *applier, struct stailq *rows) goto success; } + /* + * Do not spam WAL with excess write requests, let it process what's + * piled up first. + * This is done before opening the transaction to avoid problems with + * yielding inside it. + */ + if (journal_queue_is_full(current_journal)) + journal_wait_queue(); + /** * Explicitly begin the transaction so that we can * control fiber->gc life cycle and, in case of apply diff --git a/src/box/box.cc b/src/box/box.cc index 26cbe8aab..2b335599e 100644 --- a/src/box/box.cc +++ b/src/box/box.cc @@ -754,6 +754,28 @@ box_check_wal_mode(const char *mode_name) return (enum wal_mode) mode; } +static int64_t +box_check_wal_queue_max_len(void) +{ + int64_t len = cfg_geti64("wal_queue_max_len"); + if (len < 0) { + diag_set(ClientError, ER_CFG, "wal_queue_max_len", + "wal_queue_max_len must be >= 0"); + } + return len; +} + +static int64_t +box_check_wal_queue_max_size(void) +{ + int64_t size = cfg_geti64("wal_queue_max_size"); + if (size < 0) { + diag_set(ClientError, ER_CFG, "wal_queue_max_size", + "wal_queue_max_size must be >= 0"); + } + return size; +} + static void box_check_readahead(int readahead) { @@ -875,6 +897,10 @@ box_check_config(void) box_check_checkpoint_count(cfg_geti("checkpoint_count")); box_check_wal_max_size(cfg_geti64("wal_max_size")); box_check_wal_mode(cfg_gets("wal_mode")); + if (box_check_wal_queue_max_size() < 0) + diag_raise(); + if (box_check_wal_queue_max_len() < 0) + diag_raise(); if (box_check_memory_quota("memtx_memory") < 0) diag_raise(); box_check_memtx_min_tuple_size(cfg_geti64("memtx_min_tuple_size")); @@ -1411,6 +1437,26 @@ box_set_checkpoint_wal_threshold(void) wal_set_checkpoint_threshold(threshold); } +int +box_set_wal_queue_max_size(void) +{ + int64_t size = box_check_wal_queue_max_size(); + if (size < 0) + return -1; + wal_set_queue_max_size(size); + return 0; +} + +int +box_set_wal_queue_max_len(void) +{ + int64_t len = box_check_wal_queue_max_len(); + if (len < 0) + return -1; + wal_set_queue_max_len(len); + return 0; +} + void box_set_vinyl_memory(void) { diff --git a/src/box/box.h b/src/box/box.h index b68047a95..4f5b4b617 100644 --- a/src/box/box.h +++ b/src/box/box.h @@ -239,6 +239,8 @@ void box_set_readahead(void); void box_set_checkpoint_count(void); void box_set_checkpoint_interval(void); void box_set_checkpoint_wal_threshold(void); +int box_set_wal_queue_max_size(void); +int box_set_wal_queue_max_len(void); void box_set_memtx_memory(void); void box_set_memtx_max_tuple_size(void); void box_set_vinyl_memory(void); diff --git a/src/box/journal.c b/src/box/journal.c index cb320b557..19a184580 100644 --- a/src/box/journal.c +++ b/src/box/journal.c @@ -55,3 +55,56 @@ journal_entry_new(size_t n_rows, struct region *region, complete_data); return entry; } + +struct journal_queue_entry { + /** The fiber waiting for queue space to free. */ + struct fiber *fiber; + /** Whether the fiber should be waken up regardless of queue size. */ + bool is_ready; + /** A link in all waiting fibers list. */ + struct rlist in_queue; +}; + +/** + * Wake up the next waiter in journal queue. + */ +void +journal_queue_wakeup_next(struct rlist *link, bool force_ready) +{ + /* Empty queue or last entry in queue. */ + if (link == rlist_last(¤t_journal->waiters)) { + current_journal->queue_is_woken = false; + return; + } + /* + * When the queue isn't forcefully emptied, no need to wake everyone + * else up until there's some free space. + */ + if (journal_queue_is_full(current_journal) && !force_ready) { + current_journal->queue_is_woken = false; + return; + } + struct journal_queue_entry *e = rlist_entry(rlist_next(link), typeof(*e), + in_queue); + e->is_ready |= force_ready; + fiber_wakeup(e->fiber); +} + +void +journal_wait_queue(void) +{ + struct journal_queue_entry entry = { + .fiber = fiber(), + .is_ready = false, + }; + rlist_add_tail_entry(¤t_journal->waiters, &entry, in_queue); + /* + * Will be waken up by either queue emptying or a synchronous write. + */ + while (journal_queue_is_full(current_journal) && !entry.is_ready) + fiber_yield(); + + journal_queue_wakeup_next(&entry.in_queue, entry.is_ready); + assert(&entry.in_queue == rlist_first(¤t_journal->waiters)); + rlist_del(&entry.in_queue); +} diff --git a/src/box/journal.h b/src/box/journal.h index 5d8d5a726..9c8af062a 100644 --- a/src/box/journal.h +++ b/src/box/journal.h @@ -115,6 +115,25 @@ journal_entry_new(size_t n_rows, struct region *region, * synchronous replication. */ struct journal { + /** Maximal size of entries enqueued in journal (in bytes). */ + int64_t queue_max_size; + /** Current approximate size of journal queue. */ + int64_t queue_size; + /** Maximal allowed length of journal queue, in entries. */ + int64_t queue_max_len; + /** Current journal queue length. */ + int64_t queue_len; + /** + * The fibers waiting for some space to free in journal queue. + * Once some space is freed they will be waken up in the same order they + * entered the queue. + */ + struct rlist waiters; + /** + * Whether the queue is being woken or not. Used to avoid multiple + * concurrent wake-ups. + */ + bool queue_is_woken; /** Asynchronous write */ int (*write_async)(struct journal *journal, struct journal_entry *entry); @@ -124,6 +143,55 @@ struct journal { struct journal_entry *entry); }; +/** + * Depending on the step of recovery and instance configuration + * points at a concrete implementation of the journal. + */ +extern struct journal *current_journal; + +void +journal_queue_wakeup_next(struct rlist *link, bool force_ready); + +/** Wake the journal queue up. */ +static inline void +journal_queue_wakeup(struct journal *j, bool force_ready) +{ + assert(j == current_journal); + assert(!rlist_empty(&j->waiters)); + if (j->queue_is_woken) + return; + j->queue_is_woken = true; + journal_queue_wakeup_next(&j->waiters, force_ready); +} + +/** + * Check whether any of the queue size limits is reached. + * If the queue is full, we must wait for some of the entries to be written + * before proceeding with a new asynchronous write request. + */ +static inline bool +journal_queue_is_full(struct journal *j) +{ + assert(j == current_journal); + return (j->queue_max_size != 0 && j->queue_size >= j->queue_max_size) || + (j->queue_max_len != 0 && j->queue_len >= j->queue_max_len); +} + +/** + * Check whether anyone is waiting for the journal queue to empty. If there are + * other waiters we must go after them to preserve write order. + */ +static inline bool +journal_queue_has_waiters(struct journal *j) +{ + assert(j == current_journal); + return !rlist_empty(&j->waiters); +} + +/** Yield until there's some space in the journal queue. */ +void +journal_wait_queue(void); + /** * Complete asynchronous write. */ @@ -131,15 +199,15 @@ static inline void journal_async_complete(struct journal_entry *entry) { assert(entry->write_async_cb != NULL); + current_journal->queue_len--; + current_journal->queue_size -= entry->approx_len; + assert(current_journal->queue_len >= 0); + assert(current_journal->queue_size >= 0); + if (journal_queue_has_waiters(current_journal)) + journal_queue_wakeup(current_journal, false); entry->write_async_cb(entry); } -/** - * Depending on the step of recovery and instance configuration - * points at a concrete implementation of the journal. - */ -extern struct journal *current_journal; - /** * Write a single entry to the journal in synchronous way. * @@ -148,6 +216,17 @@ extern struct journal *current_journal; static inline int journal_write(struct journal_entry *entry) { + if (journal_queue_has_waiters(current_journal)) { + /* + * It's a synchronous write, so it's fine to wait a bit more for + * everyone else to be written. They'll wake us up back + * afterwards. + */ + journal_queue_wakeup(current_journal, true); + journal_wait_queue(); + } + current_journal->queue_size += entry->approx_len; + current_journal->queue_len += 1; return current_journal->write(current_journal, entry); } @@ -159,6 +238,12 @@ journal_write(struct journal_entry *entry) static inline int journal_write_async(struct journal_entry *entry) { + /* + * It's the job of the caller to check whether the queue is full prior + * to submitting the request. + */ + current_journal->queue_size += entry->approx_len; + current_journal->queue_len += 1; return current_journal->write_async(current_journal, entry); } @@ -198,6 +283,12 @@ journal_create(struct journal *journal, { journal->write_async = write_async; journal->write = write; + journal->queue_size = 0; + journal->queue_max_size = 0; + journal->queue_len = 0; + journal->queue_max_len = 0; + journal->queue_is_woken = false; + rlist_create(&journal->waiters); } static inline bool diff --git a/src/box/lua/cfg.cc b/src/box/lua/cfg.cc index 2d3ccbf0e..35f410710 100644 --- a/src/box/lua/cfg.cc +++ b/src/box/lua/cfg.cc @@ -164,6 +164,22 @@ lbox_cfg_set_checkpoint_wal_threshold(struct lua_State *L) return 0; } +static int +lbox_cfg_set_wal_queue_max_size(struct lua_State *L) +{ + if (box_set_wal_queue_max_size() != 0) + luaT_error(L); + return 0; +} + +static int +lbox_cfg_set_wal_queue_max_len(struct lua_State *L) +{ + if (box_set_wal_queue_max_len() != 0) + luaT_error(L); + return 0; +} + static int lbox_cfg_set_read_only(struct lua_State *L) { @@ -399,6 +415,8 @@ box_lua_cfg_init(struct lua_State *L) {"cfg_set_checkpoint_count", lbox_cfg_set_checkpoint_count}, {"cfg_set_checkpoint_interval", lbox_cfg_set_checkpoint_interval}, {"cfg_set_checkpoint_wal_threshold", lbox_cfg_set_checkpoint_wal_threshold}, + {"cfg_set_wal_queue_max_size", lbox_cfg_set_wal_queue_max_size}, + {"cfg_set_wal_queue_max_len", lbox_cfg_set_wal_queue_max_len}, {"cfg_set_read_only", lbox_cfg_set_read_only}, {"cfg_set_memtx_memory", lbox_cfg_set_memtx_memory}, {"cfg_set_memtx_max_tuple_size", lbox_cfg_set_memtx_max_tuple_size}, diff --git a/src/box/lua/load_cfg.lua b/src/box/lua/load_cfg.lua index 574c8bef4..c11a9e103 100644 --- a/src/box/lua/load_cfg.lua +++ b/src/box/lua/load_cfg.lua @@ -71,6 +71,8 @@ local default_cfg = { wal_mode = "write", wal_max_size = 256 * 1024 * 1024, wal_dir_rescan_delay= 2, + wal_queue_max_size = 0, + wal_queue_max_len = 0, force_recovery = false, replication = nil, instance_uuid = nil, @@ -163,6 +165,8 @@ local template_cfg = { coredump = 'boolean', checkpoint_interval = 'number', checkpoint_wal_threshold = 'number', + wal_queue_max_size = 'number', + wal_queue_max_len = 'number', checkpoint_count = 'number', read_only = 'boolean', hot_standby = 'boolean', @@ -277,6 +281,8 @@ local dynamic_cfg = { checkpoint_count = private.cfg_set_checkpoint_count, checkpoint_interval = private.cfg_set_checkpoint_interval, checkpoint_wal_threshold = private.cfg_set_checkpoint_wal_threshold, + wal_queue_max_size = private.cfg_set_wal_queue_max_size, + wal_queue_max_len = private.cfg_set_wal_queue_max_len, worker_pool_threads = private.cfg_set_worker_pool_threads, feedback_enabled = ifdef_feedback_set_params, feedback_crashinfo = ifdef_feedback_set_params, diff --git a/src/box/wal.c b/src/box/wal.c index 937d47ba9..9fff4220a 100644 --- a/src/box/wal.c +++ b/src/box/wal.c @@ -765,6 +765,24 @@ wal_set_checkpoint_threshold(int64_t threshold) fiber_set_cancellable(cancellable); } +void +wal_set_queue_max_size(int64_t size) +{ + struct journal *base = &wal_writer_singleton.base; + base->queue_max_size = size; + if (journal_queue_has_waiters(base) && !journal_queue_is_full(base)) + journal_queue_wakeup(base, false); +} + +void +wal_set_queue_max_len(int64_t len) +{ + struct journal *base = &wal_writer_singleton.base; + base->queue_max_len = len; + if (journal_queue_has_waiters(base) && !journal_queue_is_full(base)) + journal_queue_wakeup(base, false); +} + struct wal_gc_msg { struct cbus_call_msg base; diff --git a/src/box/wal.h b/src/box/wal.h index ca43dc6eb..1db32f66f 100644 --- a/src/box/wal.h +++ b/src/box/wal.h @@ -237,6 +237,20 @@ wal_commit_checkpoint(struct wal_checkpoint *checkpoint); void wal_set_checkpoint_threshold(int64_t threshold); +/** + * Set the pending write limit in bytes. Once the limit is reached, new + * writes are blocked until some previous writes succeed. + */ +void +wal_set_queue_max_size(int64_t size); + +/** + * Set the pending write limit in journal entries. Once the limit is reached, + * new writes are blocked until some previous writes succeeed. + */ +void +wal_set_queue_max_len(int64_t len); + /** * Remove WAL files that are not needed by consumers reading * rows at @vclock or newer. diff --git a/test/app-tap/init_script.result b/test/app-tap/init_script.result index 16c5b01d2..7a224e50e 100644 --- a/test/app-tap/init_script.result +++ b/test/app-tap/init_script.result @@ -56,6 +56,8 @@ wal_dir:. wal_dir_rescan_delay:2 wal_max_size:268435456 wal_mode:write +wal_queue_max_len:0 +wal_queue_max_size:0 worker_pool_threads:4 -- -- Test insert from detached fiber diff --git a/test/box-tap/cfg.test.lua b/test/box-tap/cfg.test.lua index a577f023d..3276ddf64 100755 --- a/test/box-tap/cfg.test.lua +++ b/test/box-tap/cfg.test.lua @@ -6,7 +6,7 @@ local socket = require('socket') local fio = require('fio') local uuid = require('uuid') local msgpack = require('msgpack') -test:plan(108) +test:plan(110) -------------------------------------------------------------------------------- -- Invalid values @@ -49,6 +49,8 @@ invalid('vinyl_run_count_per_level', 0) invalid('vinyl_run_size_ratio', 1) invalid('vinyl_bloom_fpr', 0) invalid('vinyl_bloom_fpr', 1.1) +invalid('wal_queue_max_size', -1) +invalid('wal_queue_max_len', -1) local function invalid_combinations(name, val) local status, result = pcall(box.cfg, val) diff --git a/test/box/admin.result b/test/box/admin.result index 05debe673..c818f4f9f 100644 --- a/test/box/admin.result +++ b/test/box/admin.result @@ -133,6 +133,10 @@ cfg_filter(box.cfg) - 268435456 - - wal_mode - write + - - wal_queue_max_len + - 0 + - - wal_queue_max_size + - 0 - - worker_pool_threads - 4 ... diff --git a/test/box/cfg.result b/test/box/cfg.result index 22a720c2c..19f322e7d 100644 --- a/test/box/cfg.result +++ b/test/box/cfg.result @@ -121,6 +121,10 @@ cfg_filter(box.cfg) | - 268435456 | - - wal_mode | - write + | - - wal_queue_max_len + | - 0 + | - - wal_queue_max_size + | - 0 | - - worker_pool_threads | - 4 | ... @@ -236,6 +240,10 @@ cfg_filter(box.cfg) | - 268435456 | - - wal_mode | - write + | - - wal_queue_max_len + | - 0 + | - - wal_queue_max_size + | - 0 | - - worker_pool_threads | - 4 | ... diff --git a/test/replication/gh-5536-wal-limit.result b/test/replication/gh-5536-wal-limit.result new file mode 100644 index 000000000..f7799baa8 --- /dev/null +++ b/test/replication/gh-5536-wal-limit.result @@ -0,0 +1,132 @@ +-- test-run result file version 2 +test_run = require('test_run').new() + | --- + | ... +fiber = require('fiber') + | --- + | ... + +-- +-- gh-5536: out of memory on a joining replica. Introduce a WAL queue limit so +-- that appliers stop reading new transactions from master once the queue is +-- full. +-- +box.schema.user.grant('guest', 'replication') + | --- + | ... +_ = box.schema.space.create('test') + | --- + | ... +_ = box.space.test:create_index('pk') + | --- + | ... + +replication_timeout = box.cfg.replication_timeout + | --- + | ... +box.cfg{replication_timeout=1000} + | --- + | ... + +test_run:cmd('create server replica with rpl_master=default,\ + script="replication/replica.lua"') + | --- + | - true + | ... +test_run:cmd('start server replica with wait=True, wait_load=True') + | --- + | - true + | ... + +test_run:switch('replica') + | --- + | - true + | ... +-- Huge replication timeout to not cause reconnects while applier is blocked. +-- Tiny queue size (in bytes) to allow exactly one queue entry at a time. +box.cfg{wal_queue_max_size=1, replication_timeout=1000} + | --- + | ... +write_cnt = box.error.injection.get("ERRINJ_WAL_WRITE_COUNT") + | --- + | ... +-- Block WAL writes so that we may test queue overflow. +box.error.injection.set("ERRINJ_WAL_DELAY", true) + | --- + | - ok + | ... + +test_run:switch('default') + | --- + | - true + | ... + +for i = 1,10 do box.space.test:insert{i} end + | --- + | ... + +test_run:switch('replica') + | --- + | - true + | ... +-- Wait for replication. Cannot rely on lsn bump here. It won't happen while +-- WAL is blocked. +test_run:wait_cond(function()\ + return box.error.injection.get("ERRINJ_WAL_WRITE_COUNT") > write_cnt\ +end) + | --- + | - true + | ... +require('fiber').sleep(0.5) + | --- + | ... +-- Only one entry fits when the limit is small. +assert(box.error.injection.get("ERRINJ_WAL_WRITE_COUNT") == write_cnt + 1) + | --- + | - true + | ... +box.error.injection.set("ERRINJ_WAL_DELAY", false) + | --- + | - ok + | ... + +-- Once the block is removed everything is written. +test_run:wait_cond(function()\ + return box.error.injection.get("ERRINJ_WAL_WRITE_COUNT") == write_cnt + 10\ +end) + | --- + | - true + | ... +assert(box.space.test:count() == 10) + | --- + | - true + | ... +assert(box.info.replication[1].upstream.status == 'follow') + | --- + | - true + | ... + +test_run:switch('default') + | --- + | - true + | ... + +-- Cleanup. +box.cfg{replication_timeout=replication_timeout} + | --- + | ... +test_run:cmd('stop server replica') + | --- + | - true + | ... +test_run:cmd('delete server replica') + | --- + | - true + | ... +box.space.test:drop() + | --- + | ... +box.schema.user.revoke('guest', 'replication') + | --- + | ... + diff --git a/test/replication/gh-5536-wal-limit.test.lua b/test/replication/gh-5536-wal-limit.test.lua new file mode 100644 index 000000000..1e7d61ff7 --- /dev/null +++ b/test/replication/gh-5536-wal-limit.test.lua @@ -0,0 +1,58 @@ +test_run = require('test_run').new() +fiber = require('fiber') + +-- +-- gh-5536: out of memory on a joining replica. Introduce a WAL queue limit so +-- that appliers stop reading new transactions from master once the queue is +-- full. +-- +box.schema.user.grant('guest', 'replication') +_ = box.schema.space.create('test') +_ = box.space.test:create_index('pk') + +replication_timeout = box.cfg.replication_timeout +box.cfg{replication_timeout=1000} + +test_run:cmd('create server replica with rpl_master=default,\ + script="replication/replica.lua"') +test_run:cmd('start server replica with wait=True, wait_load=True') + +test_run:switch('replica') +-- Huge replication timeout to not cause reconnects while applier is blocked. +-- Tiny queue size (in bytes) to allow exactly one queue entry at a time. +box.cfg{wal_queue_max_size=1, replication_timeout=1000} +write_cnt = box.error.injection.get("ERRINJ_WAL_WRITE_COUNT") +-- Block WAL writes so that we may test queue overflow. +box.error.injection.set("ERRINJ_WAL_DELAY", true) + +test_run:switch('default') + +for i = 1,10 do box.space.test:insert{i} end + +test_run:switch('replica') +-- Wait for replication. Cannot rely on lsn bump here. It won't happen while +-- WAL is blocked. +test_run:wait_cond(function()\ + return box.error.injection.get("ERRINJ_WAL_WRITE_COUNT") > write_cnt\ +end) +require('fiber').sleep(0.5) +-- Only one entry fits when the limit is small. +assert(box.error.injection.get("ERRINJ_WAL_WRITE_COUNT") == write_cnt + 1) +box.error.injection.set("ERRINJ_WAL_DELAY", false) + +-- Once the block is removed everything is written. +test_run:wait_cond(function()\ + return box.error.injection.get("ERRINJ_WAL_WRITE_COUNT") == write_cnt + 10\ +end) +assert(box.space.test:count() == 10) +assert(box.info.replication[1].upstream.status == 'follow') + +test_run:switch('default') + +-- Cleanup. +box.cfg{replication_timeout=replication_timeout} +test_run:cmd('stop server replica') +test_run:cmd('delete server replica') +box.space.test:drop() +box.schema.user.revoke('guest', 'replication') + diff --git a/test/replication/suite.cfg b/test/replication/suite.cfg index c80430afc..7e7004592 100644 --- a/test/replication/suite.cfg +++ b/test/replication/suite.cfg @@ -37,6 +37,7 @@ "gh-4928-tx-boundaries.test.lua": {}, "gh-5440-qsync-ro.test.lua": {}, "gh-5435-qsync-clear-synchro-queue-commit-all.test.lua": {}, + "gh-5536-wal-limit.test.lua": {}, "*": { "memtx": {"engine": "memtx"}, "vinyl": {"engine": "vinyl"} diff --git a/test/replication/suite.ini b/test/replication/suite.ini index e4812df37..89abfabff 100644 --- a/test/replication/suite.ini +++ b/test/replication/suite.ini @@ -3,7 +3,7 @@ core = tarantool script = master.lua description = tarantool/box, replication disabled = consistent.test.lua -release_disabled = catch.test.lua errinj.test.lua gc.test.lua gc_no_space.test.lua before_replace.test.lua qsync_advanced.test.lua qsync_errinj.test.lua quorum.test.lua recover_missing_xlog.test.lua sync.test.lua long_row_timeout.test.lua gh-4739-vclock-assert.test.lua gh-4730-applier-rollback.test.lua gh-5140-qsync-casc-rollback.test.lua gh-5144-qsync-dup-confirm.test.lua gh-5167-qsync-rollback-snap.test.lua gh-5506-election-on-off.test.lua +release_disabled = catch.test.lua errinj.test.lua gc.test.lua gc_no_space.test.lua before_replace.test.lua qsync_advanced.test.lua qsync_errinj.test.lua quorum.test.lua recover_missing_xlog.test.lua sync.test.lua long_row_timeout.test.lua gh-4739-vclock-assert.test.lua gh-4730-applier-rollback.test.lua gh-5140-qsync-casc-rollback.test.lua gh-5144-qsync-dup-confirm.test.lua gh-5167-qsync-rollback-snap.test.lua gh-5506-election-on-off.test.lua gh-5536-wal-limit.test.lua config = suite.cfg lua_libs = lua/fast_replica.lua lua/rlimit.lua use_unix_sockets = True -- 2.24.3 (Apple Git-128)