Re: DeadLocks...

Поиск
Список
Период
Сортировка
От Scott Marlowe
Тема Re: DeadLocks...
Дата
Msg-id 46715774.3060602@g2switchworks.com
обсуждение исходный текст
Ответ на DeadLocks...  (<tom@tacocat.net>)
Ответы Re: DeadLocks...  (<tom@tacocat.net>)
Список pgsql-general
tom@tacocat.net wrote:
> I found a problem with my application which only occurs under high loads
> (isn't that always the case?).
>
> snippets of perl...
>
> insert into tokens (token)
> select values.token
> from (values TOKEN_LIST_STRING) as values(token)
> left outer join tokens t using (token)
> where t.token_idx is null
>
> $sql =~ s/TOKEN_LIST_STRING/$string/
> where $string is of the form (('one'),('two'))
>
> This works 99% of the time.
>
> But everyone once in a long while it seems that I hit simultaneaous
> execute() statements that deadlock on the insertion.
>
> Right now I know of no other way to handle this than to eval{ } the
> execution and if it fails, sleep random milliseconds and retry... "wash
> rinse repeat" for some number of times.
>
> Is there any better way of doing this or some other means to manage
> DEADLOCK?
Is this a deadlock that postgresql detects and causes one thread to roll
back and you can recover from, or are you talking about a deadlock that
isn't detected by postgresql and locks a thread?

What error messages are you seeing?

Generally speaking, if your operations have a potential for a deadlock,
the best you can do is to do what you're doing now, detect failure and
retry x times, then give up if it won't go through.

Or, redesign the way you're doing things.

В списке pgsql-general по дате отправления:

Предыдущее
От: Tom Lane
Дата:
Сообщение: Re: strange change (and error) in 8.3 ?
Следующее
От: Greg Smith
Дата:
Сообщение: Re: PostGreSQL for a small Desktop Application