public inbox for [email protected]  
help / color / mirror / Atom feed
[PATCH] Remove PROC_IN_ANALYZE
4+ messages / 4 participants
[nested] [flat]

* [PATCH] Remove PROC_IN_ANALYZE
@ 2020-08-05 22:57  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 4+ messages in thread

From: Alvaro Herrera @ 2020-08-05 22:57 UTC (permalink / raw)

---
 src/backend/commands/analyze.c  | 13 +------------
 src/include/storage/proc.h      |  3 +--
 src/include/storage/procarray.h |  7 -------
 3 files changed, 2 insertions(+), 21 deletions(-)

diff --git a/src/backend/commands/analyze.c b/src/backend/commands/analyze.c
index 924ef37c81..e0fa73ba79 100644
--- a/src/backend/commands/analyze.c
+++ b/src/backend/commands/analyze.c
@@ -247,11 +247,8 @@ analyze_rel(Oid relid, RangeVar *relation,
 	}
 
 	/*
-	 * OK, let's do it.  First let other backends know I'm in ANALYZE.
+	 * OK, let's do it.  First, initialize progress reporting.
 	 */
-	LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE);
-	MyPgXact->vacuumFlags |= PROC_IN_ANALYZE;
-	LWLockRelease(ProcArrayLock);
 	pgstat_progress_start_command(PROGRESS_COMMAND_ANALYZE,
 								  RelationGetRelid(onerel));
 
@@ -279,14 +276,6 @@ analyze_rel(Oid relid, RangeVar *relation,
 	relation_close(onerel, NoLock);
 
 	pgstat_progress_end_command();
-
-	/*
-	 * Reset my PGXACT flag.  Note: we need this here, and not in vacuum_rel,
-	 * because the vacuum flag is cleared by the end-of-xact code.
-	 */
-	LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE);
-	MyPgXact->vacuumFlags &= ~PROC_IN_ANALYZE;
-	LWLockRelease(ProcArrayLock);
 }
 
 /*
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index b20e2ad4f6..5ceb2494ba 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -52,7 +52,6 @@ struct XidCache
  */
 #define		PROC_IS_AUTOVACUUM	0x01	/* is it an autovac worker? */
 #define		PROC_IN_VACUUM		0x02	/* currently running lazy vacuum */
-#define		PROC_IN_ANALYZE		0x04	/* currently running analyze */
 #define		PROC_VACUUM_FOR_WRAPAROUND	0x08	/* set by autovac only */
 #define		PROC_IN_LOGICAL_DECODING	0x10	/* currently doing logical
 												 * decoding outside xact */
@@ -60,7 +59,7 @@ struct XidCache
 
 /* flags reset at EOXact */
 #define		PROC_VACUUM_STATE_MASK \
-	(PROC_IN_VACUUM | PROC_IN_ANALYZE | PROC_VACUUM_FOR_WRAPAROUND)
+	(PROC_IN_VACUUM | PROC_VACUUM_FOR_WRAPAROUND)
 
 /*
  * We allow a small number of "weak" relation locks (AccessShareLock,
diff --git a/src/include/storage/procarray.h b/src/include/storage/procarray.h
index a5c7d0c064..01040d76e1 100644
--- a/src/include/storage/procarray.h
+++ b/src/include/storage/procarray.h
@@ -29,8 +29,6 @@
  */
 #define		PROCARRAY_VACUUM_FLAG			0x02	/* currently running lazy
 													 * vacuum */
-#define		PROCARRAY_ANALYZE_FLAG			0x04	/* currently running
-													 * analyze */
 #define		PROCARRAY_LOGICAL_DECODING_FLAG 0x10	/* currently doing logical
 													 * decoding outside xact */
 
@@ -42,7 +40,6 @@
  * have no corresponding PROC flag equivalent.
  */
 #define		PROCARRAY_PROC_FLAGS_MASK	(PROCARRAY_VACUUM_FLAG | \
-										 PROCARRAY_ANALYZE_FLAG | \
 										 PROCARRAY_LOGICAL_DECODING_FLAG)
 
 /* Use the following flags as an input "flags" to GetOldestXmin function */
@@ -50,10 +47,6 @@
 #define		PROCARRAY_FLAGS_DEFAULT			PROCARRAY_LOGICAL_DECODING_FLAG
 /* Ignore vacuum backends */
 #define		PROCARRAY_FLAGS_VACUUM			PROCARRAY_FLAGS_DEFAULT | PROCARRAY_VACUUM_FLAG
-/* Ignore analyze backends */
-#define		PROCARRAY_FLAGS_ANALYZE			PROCARRAY_FLAGS_DEFAULT | PROCARRAY_ANALYZE_FLAG
-/* Ignore both vacuum and analyze backends */
-#define		PROCARRAY_FLAGS_VACUUM_ANALYZE	PROCARRAY_FLAGS_DEFAULT | PROCARRAY_VACUUM_FLAG | PROCARRAY_ANALYZE_FLAG
 
 extern Size ProcArrayShmemSize(void);
 extern void CreateSharedProcArray(void);
-- 
2.20.1


--Kj7319i9nmIyA2yE--





^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* Re: do only critical work during single-user vacuum?
@ 2022-02-03 18:05  Robert Haas <[email protected]>
  0 siblings, 1 reply; 4+ messages in thread

From: Robert Haas @ 2022-02-03 18:05 UTC (permalink / raw)
  To: Andres Freund <[email protected]>; +Cc: Peter Geoghegan <[email protected]>; John Naylor <[email protected]>; PostgreSQL Hackers <[email protected]>

On Thu, Dec 9, 2021 at 8:56 PM Andres Freund <[email protected]> wrote:
> I think we should move *away* from single user mode, rather than the
> opposite. It's a substantial code burden and it's hard to use.

Yes. This thread seems to be largely devoted to the topic of making
single-user vacuum work better, but I don't see anyone asking the
question "why do we have a message that tells people to vacuum in
single user mode in the first place?". It's basically bad advice, with
one small exception that I'll talk about in a minute. Suppose we had a
message in the tree that said "HINT: Consider angering a live anaconda
to fix this problem." If that were so, the correct thing to do
wouldn't be to add a section to our documentation explaining how to
deal with angry anacondas. The correct thing to do would be to remove
the hint as bad advice that we never should have offered in the first
place. And so here. We should not try to make vacuum in single
user-mode work better or differently, or at least that shouldn't be
our primary objective. We should just stop telling people to do it. We
should probably add messages and documentation *discouraging* the use
of single user mode for recovering from wraparound trouble, exactly
the opposite of what we do now. There's nothing we can do in
single-user mode that we can't do equally well in multi-user mode. If
people try to fix wraparound problems in multi-user mode, they still
have read-only access to their database, they can use parallelism,
they can use command line utilities like vacuumdb, and they can use
psql which has line editing and allows remote access and is a way
nicer user experience than running postgres --single. We need a really
compelling reason to tell people to give up all those advantages, and
there is no such reason. It makes just as much sense as telling people
to deal with wraparound problems by angering a live anaconda.

I did say there was an exception, and it's this: the last time I
studied this issue back in 2019,[1] vacuum insisted on trying to
truncate tables even when the system is in wraparound danger. Then it
would fail, because truncating the table required allocating an XID,
which would fail if we were short on XIDs. By putting the system in
single user mode, you could continue to allocate XIDs and thus VACUUM
would work. However, if you think about this for even 10 seconds, you
can see that it's terrible. If we're so short of XIDs that we are
scared to allocate them for fear of causing an actual wraparound,
putting the system into a mode where that protection is bypassed is a
super-terrible idea. People will be able to run vacuum, yes, but if
they have too many tables, they will actually experience wraparound
and thus data loss before they process all the tables they have. What
we ought to do to solve this problem is NOT TRUNCATE when the number
of remaining XIDs is small, so that we don't consume any of the
remaining XIDs until we get the system out of wraparound danger. I
think the "failsafe" stuff Peter added in v14 fixes that, though. If
not, we should adjust it so it does. And then we should KILL WITH FIRE
the message telling people to use single user mode -- and once we do
that, the question of what the behavior ought to be when someone does
run VACUUM in single user mode becomes a lot less important.

This problem is basically self-inflicted. We have given people bad
advice (use single user mode) and then they suffer when they take it.
Ameliorating the suffering isn't the worst idea ever, but it's
basically fixing the wrong problem.

-- 
Robert Haas
EDB: http://www.enterprisedb.com

[1] http://postgr.es/m/CA+Tgmob1QCMJrHwRBK8HZtGsr+6cJANRQw2mEgJ9e=D+z7cOsw@mail.gmail.com






^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* Re: do only critical work during single-user vacuum?
@ 2022-06-27 19:36  Justin Pryzby <[email protected]>
  parent: Robert Haas <[email protected]>
  0 siblings, 1 reply; 4+ messages in thread

From: Justin Pryzby @ 2022-06-27 19:36 UTC (permalink / raw)
  To: Robert Haas <[email protected]>; +Cc: Andres Freund <[email protected]>; Peter Geoghegan <[email protected]>; John Naylor <[email protected]>; [email protected]; Hannu Krosing <[email protected]>

On Thu, Feb 03, 2022 at 01:05:50PM -0500, Robert Haas wrote:
> On Thu, Dec 9, 2021 at 8:56 PM Andres Freund <[email protected]> wrote:
> > I think we should move *away* from single user mode, rather than the
> > opposite. It's a substantial code burden and it's hard to use.
> 
> Yes. This thread seems to be largely devoted to the topic of making
> single-user vacuum work better, but I don't see anyone asking the
> question "why do we have a message that tells people to vacuum in
> single user mode in the first place?". It's basically bad advice,

> The correct thing to do would be to remove
> the hint as bad advice that we never should have offered in the first
> place. And so here. We should not try to make vacuum in single
> user-mode work better or differently, or at least that shouldn't be
> our primary objective. We should just stop telling people to do it. We
> should probably add messages and documentation *discouraging* the use
> of single user mode for recovering from wraparound trouble, exactly
> the opposite of what we do now. There's nothing we can do in
> single-user mode that we can't do equally well in multi-user mode. If
> people try to fix wraparound problems in multi-user mode, they still
> have read-only access to their database, they can use parallelism,
> they can use command line utilities like vacuumdb, and they can use
> psql which has line editing and allows remote access and is a way
> nicer user experience than running postgres --single. We need a really
> compelling reason to tell people to give up all those advantages, and
> there is no such reason. It makes just as much sense as telling people
> to deal with wraparound problems by angering a live anaconda.

By chance, I came across this prior thread which advocated the same thing in a
initially (rather than indirectly as in this year's thread).

https://www.postgresql.org/message-id/flat/CAMT0RQTmRj_Egtmre6fbiMA9E2hM3BsLULiV8W00stwa3URvzA%40mai...
|We should stop telling users to "vacuum that database in single-user mode"





^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* Re: do only critical work during single-user vacuum?
@ 2022-06-27 20:36  Peter Geoghegan <[email protected]>
  parent: Justin Pryzby <[email protected]>
  0 siblings, 0 replies; 4+ messages in thread

From: Peter Geoghegan @ 2022-06-27 20:36 UTC (permalink / raw)
  To: Justin Pryzby <[email protected]>; +Cc: Robert Haas <[email protected]>; Andres Freund <[email protected]>; John Naylor <[email protected]>; PostgreSQL Hackers <[email protected]>; Hannu Krosing <[email protected]>

On Mon, Jun 27, 2022 at 12:36 PM Justin Pryzby <[email protected]> wrote:
> By chance, I came across this prior thread which advocated the same thing in a
> initially (rather than indirectly as in this year's thread).

Revisiting this topic reminded me that PostgreSQL 14 (the first
version that had the wraparound failsafe mechanism controlled by
vacuum_failsafe_age) has been a stable release for 9 months now. As of
today I am still not aware of even one user that ran into the failsafe
mechanism in production. It might well have happened by now, of
course, but I am not aware of any specific case. Perhaps this will
change soon enough -- maybe somebody else will read this and enlighten
me.

To me the fact that the failsafe seems to seldom kick-in in practice
suggests something about workload characteristics in general: that it
isn't all that common for users to try to get away with putting off
freezing until a table attains an age that is significantly above 1
billion XIDs.

When people talk about things like 64-bit XIDs, I tend to wonder: if 2
billion XIDs wasn't enough, why should 4 billion or 8 billion be
enough? *Maybe* the system can do better by getting even further into
debt than it can today, but you can't expect to avoid freezing
altogether (without significant work elsewhere). My general sense is
that freezing isn't a particularly good thing to try to do lazily --
even if we ignore the risk of an eventual wraparound failure.

-- 
Peter Geoghegan





^ permalink  raw  reply  [nested|flat] 4+ messages in thread


end of thread, other threads:[~2022-06-27 20:36 UTC | newest]

Thread overview: 4+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2020-08-05 22:57 [PATCH] Remove PROC_IN_ANALYZE Alvaro Herrera <[email protected]>
2022-02-03 18:05 Re: do only critical work during single-user vacuum? Robert Haas <[email protected]>
2022-06-27 19:36 ` Re: do only critical work during single-user vacuum? Justin Pryzby <[email protected]>
2022-06-27 20:36   ` Re: do only critical work during single-user vacuum? Peter Geoghegan <[email protected]>

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox