Tarantool development patches archive
 help / color / mirror / Atom feed
From: Vladislav Shpilevoy <v.shpilevoy@tarantool.org>
To: Serge Petrenko <sergepetrenko@tarantool.org>,
	tarantool-patches@dev.tarantool.org
Subject: Re: [Tarantool-patches] [PATCH 4/5] [tosquash] replication: rework how local transactions wait sync
Date: Sun, 5 Jul 2020 17:13:42 +0200	[thread overview]
Message-ID: <de0f0783-da44-ed01-cc6c-ce834897bd1d@tarantool.org> (raw)
In-Reply-To: <09a2d0b1-4c7e-635a-6c9c-8a3627736ed1@tarantool.org>

>> diff --git a/src/box/txn_limbo.c b/src/box/txn_limbo.c
>> index 387cfd337..44a0c7273 100644
>> --- a/src/box/txn_limbo.c
>> +++ b/src/box/txn_limbo.c
>> @@ -316,7 +320,7 @@ txn_limbo_read_rollback(struct txn_limbo *limbo, int64_t lsn)
>>       assert(limbo->instance_id != REPLICA_ID_NIL);
>>       struct txn_limbo_entry *e, *tmp;
>>       rlist_foreach_entry_safe_reverse(e, &limbo->queue, in_queue, tmp) {
>> -        if (e->lsn <= lsn)
>> +        if (e->lsn <= lsn && txn_has_flag(e->txn, TXN_WAIT_ACK))
>>               break;
> 
> Are you rolling back the async transactions that are before the last sync
> transaction to be rolled back? Why?
> Shouldn't this condition stay the same?

Да, похоже на баг. Оставить без изменений не выйдет - у асинхронных
транзакций в лимбе лсн -1. Так что на первой асинхронной транзакции
роллбек остановится, даже если надо идти дальше. Я сделал как в
confirm в итоге. Сначала находим докуда откат. Потом делаем откат.

Тест пока не придумал. Мастер откатывает сейчас либо все, либо ничего.
Потому как откат только по таймауту, а значит всегда с самой первой.

====================
--- a/src/box/txn_limbo.c
+++ b/src/box/txn_limbo.c
@@ -319,9 +319,17 @@ txn_limbo_read_rollback(struct txn_limbo *limbo, int64_t lsn)
 {
 	assert(limbo->instance_id != REPLICA_ID_NIL);
 	struct txn_limbo_entry *e, *tmp;
-	rlist_foreach_entry_safe_reverse(e, &limbo->queue, in_queue, tmp) {
-		if (e->lsn <= lsn && txn_has_flag(e->txn, TXN_WAIT_ACK))
+	struct txn_limbo_entry *last_rollback = NULL;
+	rlist_foreach_entry_reverse(e, &limbo->queue, in_queue) {
+		if (!txn_has_flag(e->txn, TXN_WAIT_ACK))
+			continue;
+		if (e->lsn <= lsn)
 			break;
+		last_rollback = e;
+	}
+	if (last_rollback == NULL)
+		return;
+	rlist_foreach_entry_safe_reverse(e, &limbo->queue, in_queue, tmp) {
 		e->is_rollback = true;
 		txn_limbo_pop(limbo, e);
 		txn_clear_flag(e->txn, TXN_WAIT_SYNC);
@@ -342,6 +350,8 @@ txn_limbo_read_rollback(struct txn_limbo *limbo, int64_t lsn)
 			txn_complete(e->txn);
 			e->txn->fiber = fiber;
 		}
+		if (e == last_rollback)
+			break;
 	}
 }

  reply	other threads:[~2020-07-05 15:13 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-02 23:40 [Tarantool-patches] [PATCH 0/5] Qsync: local rows fixes Vladislav Shpilevoy
2020-07-02 23:40 ` [Tarantool-patches] [PATCH 1/5] [tosquash] replication: fix multiple rollbacks Vladislav Shpilevoy
2020-07-05  9:34   ` Serge Petrenko
2020-07-05 15:13     ` Vladislav Shpilevoy
2020-07-02 23:40 ` [Tarantool-patches] [PATCH 2/5] [tosquash] applier: remove unnecessary fiber name check Vladislav Shpilevoy
2020-07-05  8:40   ` Serge Petrenko
2020-07-02 23:40 ` [Tarantool-patches] [PATCH 3/5] [tosquash] txn_limbo: fix release build Vladislav Shpilevoy
2020-07-05  8:41   ` Serge Petrenko
2020-07-02 23:40 ` [Tarantool-patches] [PATCH 4/5] [tosquash] replication: rework how local transactions wait sync Vladislav Shpilevoy
2020-07-05  9:04   ` Serge Petrenko
2020-07-05 15:13     ` Vladislav Shpilevoy [this message]
2020-07-02 23:40 ` [Tarantool-patches] [PATCH 5/5] [tosquash] replication: add test on local row in the end of sync txn Vladislav Shpilevoy
2020-07-05  9:11   ` Serge Petrenko
2020-07-05 15:13     ` Vladislav Shpilevoy

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=de0f0783-da44-ed01-cc6c-ce834897bd1d@tarantool.org \
    --to=v.shpilevoy@tarantool.org \
    --cc=sergepetrenko@tarantool.org \
    --cc=tarantool-patches@dev.tarantool.org \
    --subject='Re: [Tarantool-patches] [PATCH 4/5] [tosquash] replication: rework how local transactions wait sync' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox