Guys,
This is all not about checkpoints. As I've mentioned in the first
message, even right after manual run of CHECKPOINT command in psql
pg_start_backup() takes same time (~10 minutes).
Regards,
Ivan
On Sun, Sep 28, 2008 at 8:18 PM, Simon Riggs <simon@2ndquadrant.com> wrote:
>
> On Sun, 2008-09-28 at 08:35 -0700, Joshua D. Drake wrote:
>> Ivan Zolotukhin wrote:
>> > Hello,
>> >
>> > Nothing bad both in system and postgres logs :( No serious activity
>> > during backup. I've had to change statement_timeout for backup user to
>> > make it work. But I cannot reproduce this case unfortunately.
>>
>> This is actually not uncommon and PostgreSQL shows exactly nothing in
>> terms of why it is taking so long. The only assumption I have come up
>> with is that start_backup does cause a checkpoint.
>
> Yes, it does a normal checkpoint and writes a file. No reason for it to
> take longer than any other checkpoint.
>
> At 8.2 and below checkpoints were frequently delayed on busy systems.
> This was because of lwlock starvation during commit phase of
> transactions. That was fixed in 8.3.
>
> --
> Simon Riggs www.2ndQuadrant.com
> PostgreSQL Training, Services and Support
>
>