From: Sami Imseih <samimseih@gmail.com>
Message-Id: <CDF34F36-3A52-47C8-B1D3-5D32A34EC3A7@gmail.com>
Content-Type: multipart/alternative;
	boundary="Apple-Mail=_CB260402-5A9A-480B-998C-FF6DEB5BD524"
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3774.600.62\))
Subject: Re: Restart pg_usleep when interrupted
Date: Thu, 25 Jul 2024 17:27:15 -0500
In-Reply-To: <ZpX/5/irF/gZqT7b@ip-10-97-1-34.eu-west-3.compute.internal>
Cc: Nathan Bossart <nathandbossart@gmail.com>,
 Tom Lane <tgl@sss.pgh.pa.us>,
 Robert Haas <robertmhaas@gmail.com>,
 "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
To: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
References: <Zo0UdeE3i9d0Wt5E@ip-10-97-1-34.eu-west-3.compute.internal>
 <3712E898-03DA-4E03-AF51-D3E24DE68216@gmail.com>
 <C18017A1-EDFD-4F2F-BCA1-0572D4CCC92B@gmail.com> <Zo_7blmO166ZDGbW@nathan>
 <2F781744-69B8-4BC2-A6EF-3998DCCDB634@gmail.com> <ZpA4_Pw14FCP6025@nathan>
 <01A15AEA-C35C-41DF-8E81-3B5A0B523939@gmail.com> <ZpGBjZZtbDGUtvTV@nathan>
 <521A6F3B-77E1-4F3F-B15E-BCF26D52C698@gmail.com> <ZpVaXQGEspp3gk8S@nathan>
 <ZpX/5/irF/gZqT7b@ip-10-97-1-34.eu-west-3.compute.internal>
Archived-At: 
 <https://www.postgresql.org/message-id/CDF34F36-3A52-47C8-B1D3-5D32A34EC3A7%40gmail.com>
Precedence: bulk


--Apple-Mail=_CB260402-5A9A-480B-998C-FF6DEB5BD524
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain;
	charset=utf-8

I am attaching v3 of the patch which addresses the comments made
earlier by Bertrand about the comment in the patch [1]. Also I will =
stick with
vacuum_sleep as the name as the function will be inside vacuum.c. I am =
not
sure we should make this function available outside of vacuum, but I =
would like
to hear other thoughts.

Also, earlier in the thread, Alvaro mentions what happens
if the sleep time is 0 [2]. In that case, we do not do anything as we =
check
if sleep time is > 0 microseconds before proceeding with the sleep


[1] =
https://www.postgresql.org/message-id/ZpDhS4nFX66ItAze%40ip-10-97-1-34.eu-=
west-3.compute.internal
[2] =
https://www.postgresql.org/message-id/202407120939.pr6wpjffmxov%40alvherre=
.pgsql


Regards,

Sami

=EF=BF=BC


--Apple-Mail=_CB260402-5A9A-480B-998C-FF6DEB5BD524
Content-Type: multipart/mixed;
	boundary="Apple-Mail=_7F466F60-BAB7-4FAC-AD3D-24A4B3DD3222"


--Apple-Mail=_7F466F60-BAB7-4FAC-AD3D-24A4B3DD3222
Content-Transfer-Encoding: quoted-printable
Content-Type: text/html;
	charset=us-ascii

<html><head><meta http-equiv=3D"content-type" content=3D"text/html; =
charset=3Dus-ascii"></head><body style=3D"overflow-wrap: break-word; =
-webkit-nbsp-mode: space; line-break: after-white-space;">I am attaching =
v3 of the patch which addresses the comments made<div>earlier by =
Bertrand about the comment in the patch [1]. Also I will stick =
with</div><div>vacuum_sleep as the name as the function will be inside =
vacuum.c. I am not</div><div>sure we should make this function available =
outside of vacuum, but I would like</div><div>to hear other =
thoughts.</div><div><br></div><div><div>Also, earlier in the thread, =
Alvaro mentions what happens</div><div>if the sleep time is 0 [2]. In =
that case, we do not do anything as we check</div><div>if sleep time is =
&gt; 0 microseconds before proceeding with the =
sleep</div><div><br></div></div><div style=3D"caret-color: rgb(0, 0, 0); =
color: rgb(0, 0, 0);"><br></div><div>[1]&nbsp;<a =
href=3D"https://www.postgresql.org/message-id/ZpDhS4nFX66ItAze%40ip-10-97-=
1-34.eu-west-3.compute.internal">https://www.postgresql.org/message-id/ZpD=
hS4nFX66ItAze%40ip-10-97-1-34.eu-west-3.compute.internal</a></div><div>[2]=
&nbsp;<a =
href=3D"https://www.postgresql.org/message-id/202407120939.pr6wpjffmxov%40=
alvherre.pgsql">https://www.postgresql.org/message-id/202407120939.pr6wpjf=
fmxov%40alvherre.pgsql</a></div><div><br></div><div><br></div><div>Regards=
,</div><div><br></div><div>Sami</div><div><br></div><div></div></body></ht=
ml>=

--Apple-Mail=_7F466F60-BAB7-4FAC-AD3D-24A4B3DD3222
Content-Disposition: attachment;
	filename=v3-0001-vaccum_delay-with-absolute-time-nanosleep.patch
Content-Type: application/octet-stream;
	x-unix-mode=0644;
	name="v3-0001-vaccum_delay-with-absolute-time-nanosleep.patch"
Content-Transfer-Encoding: quoted-printable

=46rom=20f77cf0f831df74e5e2d11910cea395f5f4541d25=20Mon=20Sep=2017=20=
00:00:00=202001=0AFrom:=20Sami=20Imseih=20<samimseih@gmail.com>=0ADate:=20=
Thu,=2025=20Jul=202024=2020:28:35=20+0000=0ASubject:=20[PATCH=201/1]=20=
vaccum_delay=20with=20absolute=20time=20nanosleep=0A=0A---=0A=20=
src/backend/commands/vacuum.c=20|=2049=20=
++++++++++++++++++++++++++++++++++-=0A=201=20file=20changed,=2048=20=
insertions(+),=201=20deletion(-)=0A=0Adiff=20--git=20=
a/src/backend/commands/vacuum.c=20b/src/backend/commands/vacuum.c=0A=
index=2048f8eab202..a753d6984f=20100644=0A---=20=
a/src/backend/commands/vacuum.c=0A+++=20b/src/backend/commands/vacuum.c=0A=
@@=20-116,6=20+116,53=20@@=20static=20bool=20vacuum_rel(Oid=20relid,=20=
RangeVar=20*relation,=20VacuumParams=20*params,=0A=20static=20double=20=
compute_parallel_delay(void);=0A=20static=20VacOptValue=20=
get_vacoptval_from_boolean(DefElem=20*def);=0A=20static=20bool=20=
vac_tid_reaped(ItemPointer=20itemptr,=20void=20*state);=0A+static=20void=20=
vacuum_sleep(double=20msec);=0A+=0A+static=0A+void=0A=
+vacuum_sleep(double=20msec)=0A+{=0A+=09long=09=09microsec=20=3D=20msec=20=
*=201000;=0A+=0A+=09if=20(microsec=20>=200)=0A+=09{=0A+#ifndef=20WIN32=0A=
+=09=09/*=0A+=09=09=20*=20We=20allow=20nanosleep=20to=20handle=20=
interrupts=20and=20retry=20with=20the=0A+=09=09=20*=20remaining=20time.=20=
However,=20frequent=20interruptions=20and=20restarts=20of=20the=0A+=09=09=
=20*=20nanosleep=20calls=20can=20substantially=20lead=20to=20drift=20in=20=
the=20time=20when=0A+=09=09=20*=20the=20sleep=20finally=20completes.=20=
To=20deal=20with=20this,=20we=20break=20out=20of=20the=0A+=09=09=20*=20=
loop=20whenever=20the=20elapsed=20time=20exceeds=20the=20requested=20=
time.=0A+=09=09=20*/=0A+=09=09struct=20timespec=20delay;=0A+=09=09struct=20=
timespec=20remain;=0A+=09=09struct=20instr_time=20start_time;=0A+=0A+=09=09=
INSTR_TIME_SET_CURRENT(start_time);=0A+=0A+=09=09delay.tv_sec=20=3D=20=
microsec=20/=201000000L;=0A+=09=09delay.tv_nsec=20=3D=20(microsec=20%=20=
1000000L)=20*=201000;=0A+=0A+=09=09while=20(nanosleep(&delay,=20&remain)=20=
=3D=3D=20-1=20&&=20errno=20=3D=3D=20EINTR)=0A+=09=09{=0A+=09=09=09struct=20=
instr_time=20current_time;=0A+=09=09=09struct=20instr_time=20=
elapsed_time;=0A+=0A+=09=09=09INSTR_TIME_SET_CURRENT(current_time);=0A+=0A=
+=09=09=09INSTR_TIME_SET_ZERO(elapsed_time);=0A+=09=09=09=
INSTR_TIME_ACCUM_DIFF(elapsed_time,=20current_time,=20start_time);=0A+=0A=
+=09=09=09if=20(INSTR_TIME_GET_MICROSEC(elapsed_time)=20>=3D=20microsec)=0A=
+=09=09=09=09break;=0A+=0A+=09=09=09delay=20=3D=20remain;=0A+=09=09}=0A=
+#else=0A+=09=09SleepEx((microsec=20<=20500=20?=201=20:=20(microsec=20+=20=
500)=20/=201000),=20FALSE);=0A+#endif=0A+=09}=0A+}=0A=20=0A=20/*=0A=20=20=
*=20GUC=20check=20function=20to=20ensure=20GUC=20value=20specified=20is=20=
within=20the=20allowable=0A@@=20-2384,7=20+2431,7=20@@=20=
vacuum_delay_point(void)=0A=20=09=09=09msec=20=3D=20vacuum_cost_delay=20=
*=204;=0A=20=0A=20=09=09=
pgstat_report_wait_start(WAIT_EVENT_VACUUM_DELAY);=0A-=09=09=
pg_usleep(msec=20*=201000);=0A+=09=09vacuum_sleep(msec);=0A=20=09=09=
pgstat_report_wait_end();=0A=20=0A=20=09=09/*=0A--=20=0A2.43.0=0A=0A=

--Apple-Mail=_7F466F60-BAB7-4FAC-AD3D-24A4B3DD3222
Content-Transfer-Encoding: 7bit
Content-Type: text/html;
	charset=us-ascii

<html><head><meta http-equiv="content-type" content="text/html; charset=us-ascii"></head><body style="overflow-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;"><div></div><div><br></div></body></html>
--Apple-Mail=_7F466F60-BAB7-4FAC-AD3D-24A4B3DD3222--

--Apple-Mail=_CB260402-5A9A-480B-998C-FF6DEB5BD524--