On Thu, Jun 29, 2017 at 11:39 AM, Tom Lane wrote:
> Jeff Janes writes:
> > In the now-committed version of this, the 'pg_ctl start' returns
> > successfully as soon as the server reaches a consistent state. Which is
> OK,
> > except that it does the same thing when hot_standby=off. When
> > hot
Jeff Janes writes:
> In the now-committed version of this, the 'pg_ctl start' returns
> successfully as soon as the server reaches a consistent state. Which is OK,
> except that it does the same thing when hot_standby=off. When
> hot_standby=off, I would expect it to wait for the end of recovery
On June 29, 2017 10:19:46 AM PDT, Jeff Janes wrote:
>On Tue, Jun 27, 2017 at 11:59 AM, Tom Lane wrote:
>
>> I wrote:
>> > Andres Freund writes:
>> >> On 2017-06-26 17:38:03 -0400, Tom Lane wrote:
>> >>> Hm. Take that a bit further, and we could drop the connection
>probes
>> >>> altogether --
On Tue, Jun 27, 2017 at 11:59 AM, Tom Lane wrote:
> I wrote:
> > Andres Freund writes:
> >> On 2017-06-26 17:38:03 -0400, Tom Lane wrote:
> >>> Hm. Take that a bit further, and we could drop the connection probes
> >>> altogether --- just put the whole responsibility on the postmaster to
> >>>
On Wed, Jun 28, 2017 at 8:31 PM, Tom Lane wrote:
> Andres Freund writes:
>> On 2017-06-27 14:59:18 -0400, Tom Lane wrote:
>>> However, it's certainly arguable that this is too much change for an
>>> optional post-beta patch.
>
>> Yea, I think there's a valid case to be made for that. I'm still
>>
Alvaro Herrera writes:
> Tom Lane wrote:
>> So when I removed the miscadmin.h include, I found out that pg_ctl is
>> also relying on PG_BACKEND_VERSIONSTR from that file.
>>
>> There are at least three things we could do here:
>>
>> 1. Give this up as not worth this much trouble.
>>
>> 2. Move
Tom Lane wrote:
> Andres Freund writes:
> So when I removed the miscadmin.h include, I found out that pg_ctl is
> also relying on PG_BACKEND_VERSIONSTR from that file.
>
> There are at least three things we could do here:
>
> 1. Give this up as not worth this much trouble.
>
> 2. Move PG_BACKE
Andres Freund writes:
> On 2017-06-28 13:31:27 -0400, Tom Lane wrote:
>> While looking this over again, I got worried about the fact that pg_ctl
>> is #including "miscadmin.h". That's a pretty low-level backend header
>> and it wouldn't be surprising at all if somebody tried to put stuff in
>> it
On 2017-06-28 13:31:27 -0400, Tom Lane wrote:
> I'm not hearing anyone speaking against doing this now, so I'm going
> to go ahead with it.
Cool.
> While looking this over again, I got worried about the fact that pg_ctl
> is #including "miscadmin.h". That's a pretty low-level backend header
> a
Andres Freund writes:
> On 2017-06-27 14:59:18 -0400, Tom Lane wrote:
>> However, it's certainly arguable that this is too much change for an
>> optional post-beta patch.
> Yea, I think there's a valid case to be made for that. I'm still
> inclined to go along with this, it seems we're otherwise
Andres Freund writes:
> On 2017-06-27 14:59:18 -0400, Tom Lane wrote:
>> If we decide that it has to wait for v11,
>> I'd address Jeff's complaint by hacking the loop behavior in
>> test_postmaster_connection, which'd be ugly but not many lines of code.
> Basically increasing the wait time over t
Hi,
On 2017-06-27 14:59:18 -0400, Tom Lane wrote:
> Here's a draft patch for that. I quite like the results --- this seems
> way simpler and more reliable than what pg_ctl has done up to now.
Yea, I like that too.
> However, it's certainly arguable that this is too much change for an
> optiona
I wrote:
> Andres Freund writes:
>> On 2017-06-26 17:38:03 -0400, Tom Lane wrote:
>>> Hm. Take that a bit further, and we could drop the connection probes
>>> altogether --- just put the whole responsibility on the postmaster to
>>> show in the pidfile whether it's ready for connections or not.
Andres Freund writes:
> On 2017-06-26 17:38:03 -0400, Tom Lane wrote:
>> Hm. Take that a bit further, and we could drop the connection probes
>> altogether --- just put the whole responsibility on the postmaster to
>> show in the pidfile whether it's ready for connections or not.
> Yea, that see
On 2017-06-26 17:38:03 -0400, Tom Lane wrote:
> Andres Freund writes:
> > On 2017-06-26 17:30:30 -0400, Tom Lane wrote:
> >> No, I don't like that at all. Has race conditions against updates
> >> coming from the startup process.
>
> > You'd obviously have to take the appropriate locks. I think
Andres Freund writes:
> On 2017-06-26 17:30:30 -0400, Tom Lane wrote:
>> No, I don't like that at all. Has race conditions against updates
>> coming from the startup process.
> You'd obviously have to take the appropriate locks. I think the issue
> here is less race conditions, and more that ar
On 2017-06-26 17:30:30 -0400, Tom Lane wrote:
> Andres Freund writes:
> > It'd be quite possible to address the race-condition by moving the
> > updating of the control file to postmaster, to the
> > CheckPostmasterSignal(PMSIGNAL_BEGIN_HOT_STANDBY) block. That'd require
> > updating the control f
Andres Freund writes:
> It'd be quite possible to address the race-condition by moving the
> updating of the control file to postmaster, to the
> CheckPostmasterSignal(PMSIGNAL_BEGIN_HOT_STANDBY) block. That'd require
> updating the control file from postmaster, which'd be somewhat ugly.
No, I do
Hi,
On 2017-06-26 16:49:07 -0400, Tom Lane wrote:
> Andres Freund writes:
> > Arguably we could and should improve the logic when the server has
> > started, right now it's pretty messy because we never treat a standby as
> > up if hot_standby is disabled...
>
> True. If you could tell the diff
Andres Freund writes:
> Arguably we could and should improve the logic when the server has
> started, right now it's pretty messy because we never treat a standby as
> up if hot_standby is disabled...
True. If you could tell the difference between "HS disabled" and "HS not
enabled yet" from pg_c
I wrote:
> Andres Freund writes:
>> It'd not be unreasonble to check pg_control first, and only after that
>> indicates readyness check via the protocol.
> Hm, that's a thought. The problem here isn't the frequency of checks,
> but the log spam.
Actually, that wouldn't help much as things stand
On 2017-06-26 16:26:00 -0400, Tom Lane wrote:
> Andres Freund writes:
> > On 2017-06-26 16:19:16 -0400, Tom Lane wrote:
> >> Sure, what do you think an appropriate behavior would be?
>
> > It'd not be unreasonble to check pg_control first, and only after that
> > indicates readyness check via the
Andres Freund writes:
> On 2017-06-26 16:19:16 -0400, Tom Lane wrote:
>> Sure, what do you think an appropriate behavior would be?
> It'd not be unreasonble to check pg_control first, and only after that
> indicates readyness check via the protocol.
Hm, that's a thought. The problem here isn't
On 2017-06-26 16:19:16 -0400, Tom Lane wrote:
> Jeff Janes writes:
> > The 10 fold increase in log spam during long PITR recoveries is a bit
> > unfortunate.
>
> > 9153 2017-06-26 12:55:40.243 PDT FATAL: the database system is starting up
> > 9154 2017-06-26 12:55:40.345 PDT FATAL: the databa
Jeff Janes writes:
> The 10 fold increase in log spam during long PITR recoveries is a bit
> unfortunate.
> 9153 2017-06-26 12:55:40.243 PDT FATAL: the database system is starting up
> 9154 2017-06-26 12:55:40.345 PDT FATAL: the database system is starting up
> 9156 2017-06-26 12:55:40.447 P
On Mon, Jun 26, 2017 at 12:15 PM, Tom Lane wrote:
> Michael Paquier writes:
> > On Mon, Jun 26, 2017 at 7:13 AM, Tom Lane wrote:
> >> The attached proposed patch adjusts pg_ctl to check every 100msec,
> >> instead of every second, for the postmaster to be done starting or
> >> stopping.
>
> >>
Michael Paquier writes:
> On Mon, Jun 26, 2017 at 7:13 AM, Tom Lane wrote:
>> The attached proposed patch adjusts pg_ctl to check every 100msec,
>> instead of every second, for the postmaster to be done starting or
>> stopping.
>> +#define WAITS_PER_SEC 10 /* should divide 100 evenly *
On Mon, Jun 26, 2017 at 7:13 AM, Tom Lane wrote:
> The attached proposed patch adjusts pg_ctl to check every 100msec,
> instead of every second, for the postmaster to be done starting or
> stopping. This cuts the runtime of the recovery TAP tests from around
> 4m30s to around 3m10s on my machine,
I still have a bee in my bonnet about how slow the recovery TAP tests
are, and especially about how low the CPU usage is while they run,
suggesting that a lot of the wall clock time is being expended on
useless sleeps. Some analysis I did today found some low-hanging fruit
there: a significant par
29 matches
Mail list logo