Antw: Performance problem with query

On 7/18/06, Christian Rengstl
<Christian.Rengstl@klinik.uni-regensburg.de> wrote:

now finally after a long time i have the query plan for the whole filled table. I hope somebody can explain me why it takes so much longer...
QUERY PLAN

--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
-----
Result (cost=0.09..254643.09 rows=8044000 width=39) (actual time=0.043..0.043
rows=0 loops=1)
One-Time Filter: ((((split_part(($1)::text, '_'::text, 2))::smallint = 1) IS
NOT TRUE) AND (((split_part(($2)::text, '_'::text, 2))::smallint = 22) IS NOT TR
UE))
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (never executed)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (never executed)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual time=0.006..0.006 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (never executed)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (never executed)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=39) (nev
er executed)
Total runtime: 0.238 ms

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual time=0.056..655772
.273 rows=8044000 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text, 2))::smallint = 1)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual time=0.003..0.003 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (actual time=0.002..0.002 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual time=0.006..0.007 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=39) (act
ual time=0.002..191672.344 rows=8044000 loops=1)
Total runtime: 62259544.896 ms

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual time=1.245..1.245
rows=0 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text, 2))::smallint = 22)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (never executed)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (never executed)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual time=0.013..0.014 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (actual time=0.009..0.009 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=39) (nev
er executed)
Total runtime: 22.270 ms
(31 Zeilen)

"Christian Rengstl" <Christian.Rengstl@klinik.uni-regensburg.de> 13.07.06 8.37 Uhr >>>

Good morning list,

the following query takes about 15 to 20 minutes for around 2 million lines in the file myfile.txt, but with 8 million lines it takes around 5 hours and i just don't understand why there is such a huge discrepancy in performance.

COPY public.temp_table FROM 'myfile.txt' WITH DELIMITER '\t';

INSERT INTO public.master(pid,smid, val1, val2, chr)
SELECT pid, smid, val1, val12, CAST(split_part((SELECT chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

INSERT INTO public.values(smid, pos, chr)
SELECT DISTINCT smid, pos, CAST(split_part((SELECT chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

what is this phrase doing exactly?
CAST(split_part((SELECT chr from public.temp_table LIMIT 1), '_', 2)

it looks fishy.
merlin

On 19/07/2006, at 4:24 AM, Christian Rengstl wrote:

now finally after a long time i have the query plan for the whole
filled table. I hope somebody can explain me why it takes so much
longer...

These explain analyze results don't appear to be from the queries you
posted previously. For these results to mean anything you need to
include the EXACT queries you used to generate them.

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual
time=0.056..655772
.273 rows=8044000 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text,
2))::smallint = 1)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.003..0.003 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.002..0.002 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.006..0.007 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000
width=39) (act
ual time=0.002..191672.344 rows=8044000 loops=1)
Total runtime: 62259544.896 ms

This is the query you want to be interested in, the others took no
time at all.

As a guess I would say the query is an INSERT INTO ... FROM
SELECT ... WHERE (split_part(???, '_', 2))::smallint = 1 statement.
The majority of the time appears to be taken up on the actual INSERT
and not the SELECT part.

How many rows are in the target table and what indexes does it have?
Does it have any triggers, check constraints, or rules applied to it?
All these things can make the insert take longer as the number of
rows you have already in the table increases.

More than likely you have a target table with a LOT of rows and a
bunch of indexes on it and your disks are being thrashed because the
indexes are not able to stay cached in RAM. At this point you should
ensure your machine is not swapping do disk, and at the very least
you should go through one of the many tuning guidelines available and
ensure you have allocated the appropriate amount of memory to
postgresql for your needs.

You may also want to consider dropping the indexes before you do the
INSERT and recreate them afterwards.

"Christian Rengstl" <Christian.Rengstl@klinik.uni-regensburg.de>
13.07.06 8.37 Uhr >>>

Good morning list,

the following query takes about 15 to 20 minutes for around 2
million lines in the file myfile.txt, but with 8 million lines it
takes around 5 hours and i just don't understand why there is such
a huge discrepancy in performance.

COPY public.temp_table FROM 'myfile.txt' WITH DELIMITER '\t';

INSERT INTO public.master(pid,smid, val1, val2, chr)
SELECT pid, smid, val1, val12, CAST(split_part((SELECT
chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

INSERT INTO public.values(smid, pos, chr)
SELECT DISTINCT smid, pos, CAST(split_part((SELECT chr from
public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

I came up with this query, because i wanted to use the COPY command
to load huge files into the db, but i don't want to load all the
columns contained in the file in only one table but copy some of
them into one table and some in a second table. As i found out with
logging, the data is loaded into temp_table within 15 minutes, but
to transfer it from the temp_table toagain only something like 10
minutes. Can it be that the cast takes up so much more time than
when reading and transferring 2 million lines?

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

The thing is that in the text file there is a column that is something like xyz_12 and in the table i just need the integer part of it that's what the query is used for. The problem though is not really in the select part, because running the select part on 8 million lines takes about 3 minutes, but i don't know why the insert is taking so long.

"Merlin Moncure" <mmoncure@gmail.com> wrote on 07/18/06 9:39 pm:

On 7/18/06, Christian Rengstl
<Christian.Rengstl@klinik.uni-regensburg.de> wrote:

now finally after a long time i have the query plan for the whole filled table. I hope somebody can explain me why it takes so much longer...
QUERY PLAN

--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
-----
Result (cost=0.09..254643.09 rows=8044000 width=39) (actual time=0.043..0.043
rows=0 loops=1)
One-Time Filter: ((((split_part(($1)::text, '_'::text, 2))::smallint = 1) IS
NOT TRUE) AND (((split_part(($2)::text, '_'::text, 2))::smallint = 22) IS NOT TR
UE))
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (never executed)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (never executed)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual time=0.006..0.006 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (never executed)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (never executed)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=39) (nev
er executed)
Total runtime: 0.238 ms

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual time=0.056..655772
.273 rows=8044000 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text, 2))::smallint = 1)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual time=0.003..0.003 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (actual time=0.002..0.002 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual time=0.006..0.007 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=39) (act
ual time=0.002..191672.344 rows=8044000 loops=1)
Total runtime: 62259544.896 ms

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual time=1.245..1.245
rows=0 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text, 2))::smallint = 22)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (never executed)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (never executed)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual time=0.013..0.014 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=
10) (actual time=0.009..0.009 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000 width=39) (nev
er executed)
Total runtime: 22.270 ms
(31 Zeilen)

"Christian Rengstl" <Christian.Rengstl@klinik.uni-regensburg.de> 13.07.06 8.37 Uhr >>>

Good morning list,

the following query takes about 15 to 20 minutes for around 2 million lines in the file myfile.txt, but with 8 million lines it takes around 5 hours and i just don't understand why there is such a huge discrepancy in performance.

COPY public.temp_table FROM 'myfile.txt' WITH DELIMITER '\t';

INSERT INTO public.master(pid,smid, val1, val2, chr)
SELECT pid, smid, val1, val12, CAST(split_part((SELECT chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

INSERT INTO public.values(smid, pos, chr)
SELECT DISTINCT smid, pos, CAST(split_part((SELECT chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

what is this phrase doing exactly?
CAST(split_part((SELECT chr from public.temp_table LIMIT 1), '_', 2)

it looks fishy.
merlin

---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

The analyze is from the exact query and i dropped the indexes before the insert as well without imrpvement. The target table is as well completely empty and the insert is supposed to write, in this case, more or less 8 million lines in the table. There is a rule though, because i have inheritance table structure with one master table and around 20 child tables.

Q <qdolan@gmail.com> wrote on 07/19/06 4:37 am:

On 19/07/2006, at 4:24 AM, Christian Rengstl wrote:

now finally after a long time i have the query plan for the whole
filled table. I hope somebody can explain me why it takes so much
longer...

These explain analyze results don't appear to be from the queries you
posted previously. For these results to mean anything you need to
include the EXACT queries you used to generate them.

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual
time=0.056..655772
.273 rows=8044000 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text,
2))::smallint = 1)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.003..0.003 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.002..0.002 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.006..0.007 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000
width=39) (act
ual time=0.002..191672.344 rows=8044000 loops=1)
Total runtime: 62259544.896 ms

This is the query you want to be interested in, the others took no
time at all.

As a guess I would say the query is an INSERT INTO ... FROM
SELECT ... WHERE (split_part(???, '_', 2))::smallint = 1 statement.
The majority of the time appears to be taken up on the actual INSERT
and not the SELECT part.

How many rows are in the target table and what indexes does it have?
Does it have any triggers, check constraints, or rules applied to it?
All these things can make the insert take longer as the number of
rows you have already in the table increases.

More than likely you have a target table with a LOT of rows and a
bunch of indexes on it and your disks are being thrashed because the
indexes are not able to stay cached in RAM. At this point you should
ensure your machine is not swapping do disk, and at the very least
you should go through one of the many tuning guidelines available and
ensure you have allocated the appropriate amount of memory to
postgresql for your needs.

You may also want to consider dropping the indexes before you do the
INSERT and recreate them afterwards.

"Christian Rengstl" <Christian.Rengstl@klinik.uni-regensburg.de>
13.07.06 8.37 Uhr >>>

Good morning list,

the following query takes about 15 to 20 minutes for around 2
million lines in the file myfile.txt, but with 8 million lines it
takes around 5 hours and i just don't understand why there is such
a huge discrepancy in performance.

COPY public.temp_table FROM 'myfile.txt' WITH DELIMITER '\t';

INSERT INTO public.master(pid,smid, val1, val2, chr)
SELECT pid, smid, val1, val12, CAST(split_part((SELECT
chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

INSERT INTO public.values(smid, pos, chr)
SELECT DISTINCT smid, pos, CAST(split_part((SELECT chr from
public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

I came up with this query, because i wanted to use the COPY command
to load huge files into the db, but i don't want to load all the
columns contained in the file in only one table but copy some of
them into one table and some in a second table. As i found out with
logging, the data is loaded into temp_table within 15 minutes, but
to transfer it from the temp_table toagain only something like 10
minutes. Can it be that the cast takes up so much more time than
when reading and transferring 2 million lines?

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

---------------------------(end of broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

On 19/07/2006, at 6:32 PM, Christian Rengstl wrote:

The analyze is from the exact query and i dropped the indexes
before the insert as well without imrpvement. The target table is
as well completely empty and the insert is supposed to write, in
this case, more or less 8 million lines in the table. There is a
rule though, because i have inheritance table structure with one
master table and around 20 child tables.

I would say the problem is in the rule. Try doing the insert into a
duplicate table with no rules or inheritance and see how long it takes.

Perhaps you should provide the actual schema of tables and rules that
are involved in the query in question.

Q <qdolan@gmail.com> wrote on 07/19/06 4:37 am:

On 19/07/2006, at 4:24 AM, Christian Rengstl wrote:

now finally after a long time i have the query plan for the whole
filled table. I hope somebody can explain me why it takes so much
longer...

These explain analyze results don't appear to be from the queries you
posted previously. For these results to mean anything you need to
include the EXACT queries you used to generate them.

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual
time=0.056..655772
.273 rows=8044000 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text,
2))::smallint = 1)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.003..0.003 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.002..0.002 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.006..0.007 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000
width=39) (act
ual time=0.002..191672.344 rows=8044000 loops=1)
Total runtime: 62259544.896 ms

This is the query you want to be interested in, the others took no
time at all.

As a guess I would say the query is an INSERT INTO ... FROM
SELECT ... WHERE (split_part(???, '_', 2))::smallint = 1 statement.
The majority of the time appears to be taken up on the actual INSERT
and not the SELECT part.

How many rows are in the target table and what indexes does it have?
Does it have any triggers, check constraints, or rules applied to it?
All these things can make the insert take longer as the number of
rows you have already in the table increases.

More than likely you have a target table with a LOT of rows and a
bunch of indexes on it and your disks are being thrashed because the
indexes are not able to stay cached in RAM. At this point you should
ensure your machine is not swapping do disk, and at the very least
you should go through one of the many tuning guidelines available and
ensure you have allocated the appropriate amount of memory to
postgresql for your needs.

You may also want to consider dropping the indexes before you do the
INSERT and recreate them afterwards.

"Christian Rengstl" <Christian.Rengstl@klinik.uni-regensburg.de>
13.07.06 8.37 Uhr >>>

Good morning list,

the following query takes about 15 to 20 minutes for around 2
million lines in the file myfile.txt, but with 8 million lines it
takes around 5 hours and i just don't understand why there is such
a huge discrepancy in performance.

COPY public.temp_table FROM 'myfile.txt' WITH DELIMITER '\t';

INSERT INTO public.master(pid,smid, val1, val2, chr)
SELECT pid, smid, val1, val12, CAST(split_part((SELECT
chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

INSERT INTO public.values(smid, pos, chr)
SELECT DISTINCT smid, pos, CAST(split_part((SELECT chr from
public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

I came up with this query, because i wanted to use the COPY command
to load huge files into the db, but i don't want to load all the
columns contained in the file in only one table but copy some of
them into one table and some in a second table. As i found out with
logging, the data is loaded into temp_table within 15 minutes, but
to transfer it from the temp_table toagain only something like 10
minutes. Can it be that the cast takes up so much more time than
when reading and transferring 2 million lines?

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

---------------------------(end of
broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

Well, i'll try that, but honestly i don't think it's the rule as the rule is really simple: it's just one simple integer comparison...

Q <qdolan@gmail.com> wrote on 07/19/06 11:54 am:

On 19/07/2006, at 6:32 PM, Christian Rengstl wrote:

The analyze is from the exact query and i dropped the indexes
before the insert as well without imrpvement. The target table is
as well completely empty and the insert is supposed to write, in
this case, more or less 8 million lines in the table. There is a
rule though, because i have inheritance table structure with one
master table and around 20 child tables.

I would say the problem is in the rule. Try doing the insert into a
duplicate table with no rules or inheritance and see how long it takes.

Perhaps you should provide the actual schema of tables and rules that
are involved in the query in question.

Q <qdolan@gmail.com> wrote on 07/19/06 4:37 am:

On 19/07/2006, at 4:24 AM, Christian Rengstl wrote:

now finally after a long time i have the query plan for the whole
filled table. I hope somebody can explain me why it takes so much
longer...

These explain analyze results don't appear to be from the queries you
posted previously. For these results to mean anything you need to
include the EXACT queries you used to generate them.

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual
time=0.056..655772
.273 rows=8044000 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text,
2))::smallint = 1)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.003..0.003 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.002..0.002 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.006..0.007 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000
width=39) (act
ual time=0.002..191672.344 rows=8044000 loops=1)
Total runtime: 62259544.896 ms

This is the query you want to be interested in, the others took no
time at all.

As a guess I would say the query is an INSERT INTO ... FROM
SELECT ... WHERE (split_part(???, '_', 2))::smallint = 1 statement.
The majority of the time appears to be taken up on the actual INSERT
and not the SELECT part.

How many rows are in the target table and what indexes does it have?
Does it have any triggers, check constraints, or rules applied to it?
All these things can make the insert take longer as the number of
rows you have already in the table increases.

More than likely you have a target table with a LOT of rows and a
bunch of indexes on it and your disks are being thrashed because the
indexes are not able to stay cached in RAM. At this point you should
ensure your machine is not swapping do disk, and at the very least
you should go through one of the many tuning guidelines available and
ensure you have allocated the appropriate amount of memory to
postgresql for your needs.

You may also want to consider dropping the indexes before you do the
INSERT and recreate them afterwards.

"Christian Rengstl" <Christian.Rengstl@klinik.uni-regensburg.de>
13.07.06 8.37 Uhr >>>

Good morning list,

the following query takes about 15 to 20 minutes for around 2
million lines in the file myfile.txt, but with 8 million lines it
takes around 5 hours and i just don't understand why there is such
a huge discrepancy in performance.

COPY public.temp_table FROM 'myfile.txt' WITH DELIMITER '\t';

INSERT INTO public.master(pid,smid, val1, val2, chr)
SELECT pid, smid, val1, val12, CAST(split_part((SELECT
chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

INSERT INTO public.values(smid, pos, chr)
SELECT DISTINCT smid, pos, CAST(split_part((SELECT chr from
public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

I came up with this query, because i wanted to use the COPY command
to load huge files into the db, but i don't want to load all the
columns contained in the file in only one table but copy some of
them into one table and some in a second table. As i found out with
logging, the data is loaded into temp_table within 15 minutes, but
to transfer it from the temp_table toagain only something like 10
minutes. Can it be that the cast takes up so much more time than
when reading and transferring 2 million lines?

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

---------------------------(end of
broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

Obviously it had something to do with the rule, because now everything finished within 20 minutes. the problem is just that i don't really want to give up the inheritance design. is there a way to maintain the inheritance that doesn't cause this huge performance problem?

Q <qdolan@gmail.com> wrote on 07/19/06 11:54 am:

On 19/07/2006, at 6:32 PM, Christian Rengstl wrote:

The analyze is from the exact query and i dropped the indexes
before the insert as well without imrpvement. The target table is
as well completely empty and the insert is supposed to write, in
this case, more or less 8 million lines in the table. There is a
rule though, because i have inheritance table structure with one
master table and around 20 child tables.

I would say the problem is in the rule. Try doing the insert into a
duplicate table with no rules or inheritance and see how long it takes.

Perhaps you should provide the actual schema of tables and rules that
are involved in the query in question.

Q <qdolan@gmail.com> wrote on 07/19/06 4:37 am:

On 19/07/2006, at 4:24 AM, Christian Rengstl wrote:

now finally after a long time i have the query plan for the whole
filled table. I hope somebody can explain me why it takes so much
longer...

These explain analyze results don't appear to be from the queries you
posted previously. For these results to mean anything you need to
include the EXACT queries you used to generate them.

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual
time=0.056..655772
.273 rows=8044000 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text,
2))::smallint = 1)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.003..0.003 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.002..0.002 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.006..0.007 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000
width=39) (act
ual time=0.002..191672.344 rows=8044000 loops=1)
Total runtime: 62259544.896 ms

This is the query you want to be interested in, the others took no
time at all.

As a guess I would say the query is an INSERT INTO ... FROM
SELECT ... WHERE (split_part(???, '_', 2))::smallint = 1 statement.
The majority of the time appears to be taken up on the actual INSERT
and not the SELECT part.

How many rows are in the target table and what indexes does it have?
Does it have any triggers, check constraints, or rules applied to it?
All these things can make the insert take longer as the number of
rows you have already in the table increases.

More than likely you have a target table with a LOT of rows and a
bunch of indexes on it and your disks are being thrashed because the
indexes are not able to stay cached in RAM. At this point you should
ensure your machine is not swapping do disk, and at the very least
you should go through one of the many tuning guidelines available and
ensure you have allocated the appropriate amount of memory to
postgresql for your needs.

You may also want to consider dropping the indexes before you do the
INSERT and recreate them afterwards.

"Christian Rengstl" <Christian.Rengstl@klinik.uni-regensburg.de>
13.07.06 8.37 Uhr >>>

Good morning list,

the following query takes about 15 to 20 minutes for around 2
million lines in the file myfile.txt, but with 8 million lines it
takes around 5 hours and i just don't understand why there is such
a huge discrepancy in performance.

COPY public.temp_table FROM 'myfile.txt' WITH DELIMITER '\t';

INSERT INTO public.master(pid,smid, val1, val2, chr)
SELECT pid, smid, val1, val12, CAST(split_part((SELECT
chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

INSERT INTO public.values(smid, pos, chr)
SELECT DISTINCT smid, pos, CAST(split_part((SELECT chr from
public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

I came up with this query, because i wanted to use the COPY command
to load huge files into the db, but i don't want to load all the
columns contained in the file in only one table but copy some of
them into one table and some in a second table. As i found out with
logging, the data is loaded into temp_table within 15 minutes, but
to transfer it from the temp_table toagain only something like 10
minutes. Can it be that the cast takes up so much more time than
when reading and transferring 2 million lines?

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

---------------------------(end of
broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

On 19/07/2006, at 8:49 PM, Christian Rengstl wrote:

Obviously it had something to do with the rule, because now
everything finished within 20 minutes. the problem is just that i
don't really want to give up the inheritance design. is there a way
to maintain the inheritance that doesn't cause this huge
performance problem?

That is hard to say unless you post the rule and table schema you are
currently using.

Q <qdolan@gmail.com> wrote on 07/19/06 11:54 am:

On 19/07/2006, at 6:32 PM, Christian Rengstl wrote:

The analyze is from the exact query and i dropped the indexes
before the insert as well without imrpvement. The target table is
as well completely empty and the insert is supposed to write, in
this case, more or less 8 million lines in the table. There is a
rule though, because i have inheritance table structure with one
master table and around 20 child tables.

I would say the problem is in the rule. Try doing the insert into a
duplicate table with no rules or inheritance and see how long it
takes.

Perhaps you should provide the actual schema of tables and rules that
are involved in the query in question.

Q <qdolan@gmail.com> wrote on 07/19/06 4:37 am:

On 19/07/2006, at 4:24 AM, Christian Rengstl wrote:

now finally after a long time i have the query plan for the whole
filled table. I hope somebody can explain me why it takes so much
longer...

These explain analyze results don't appear to be from the
queries you
posted previously. For these results to mean anything you need to
include the EXACT queries you used to generate them.

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual
time=0.056..655772
.273 rows=8044000 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text,
2))::smallint = 1)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.003..0.003 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.002..0.002 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.006..0.007 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000
width=39) (act
ual time=0.002..191672.344 rows=8044000 loops=1)
Total runtime: 62259544.896 ms

This is the query you want to be interested in, the others took no
time at all.

As a guess I would say the query is an INSERT INTO ... FROM
SELECT ... WHERE (split_part(???, '_', 2))::smallint = 1 statement.
The majority of the time appears to be taken up on the actual
INSERT
and not the SELECT part.

How many rows are in the target table and what indexes does it
have?
Does it have any triggers, check constraints, or rules applied
to it?
All these things can make the insert take longer as the number of
rows you have already in the table increases.

More than likely you have a target table with a LOT of rows and a
bunch of indexes on it and your disks are being thrashed because
the
indexes are not able to stay cached in RAM. At this point you
should
ensure your machine is not swapping do disk, and at the very least
you should go through one of the many tuning guidelines
available and
ensure you have allocated the appropriate amount of memory to
postgresql for your needs.

You may also want to consider dropping the indexes before you do
the
INSERT and recreate them afterwards.

"Christian Rengstl" <Christian.Rengstl@klinik.uni-
regensburg.de>
13.07.06 8.37 Uhr >>>

Good morning list,

the following query takes about 15 to 20 minutes for around 2
million lines in the file myfile.txt, but with 8 million lines it
takes around 5 hours and i just don't understand why there is such
a huge discrepancy in performance.

COPY public.temp_table FROM 'myfile.txt' WITH DELIMITER '\t';

INSERT INTO public.master(pid,smid, val1, val2, chr)
SELECT pid, smid, val1, val12, CAST(split_part((SELECT
chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

INSERT INTO public.values(smid, pos, chr)
SELECT DISTINCT smid, pos, CAST(split_part((SELECT chr
from
public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

I came up with this query, because i wanted to use the COPY
command
to load huge files into the db, but i don't want to load all the
columns contained in the file in only one table but copy some of
them into one table and some in a second table. As i found out
with
logging, the data is loaded into temp_table within 15 minutes, but
to transfer it from the temp_table toagain only something like 10
minutes. Can it be that the cast takes up so much more time than
when reading and transferring 2 million lines?

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan -
qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

---------------------------(end of
broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

So here's the master table including the rules:

entry_no int8 NOT NULL DEFAULT nextval('public.master_seq'::regclass),
pid varchar(15) NOT NULL,
val_1 varchar(1),
val_2 varchar(1),
chr int2 NOT NULL,
aendat timestamp DEFAULT now(),
aennam varchar(8),
CONSTRAINT "PK_ENTRY" PRIMARY KEY (entry_no),
CONSTRAINT "UNIQUE_MASTER" UNIQUE (pid, entry_no)

CREATE OR REPLACE RULE "INSERT_INTO_1" AS
ON INSERT TO public.master
WHERE new.chr = 1 DO INSTEAD INSERT INTO public.table_1 (entry_no, pid, val_1, val_2, chr, aendat, aennam)
VALUES (new.entry_no, new.pid, new.val_1, new.val_2, new.chr, new.aendat, new.aennam);

Like this i have around 20 rules so far, but there might be more later on. The children tables are so far exactly as the master table.

Q <qdolan@gmail.com> wrote on 07/19/06 1:52 pm:

On 19/07/2006, at 8:49 PM, Christian Rengstl wrote:

Obviously it had something to do with the rule, because now
everything finished within 20 minutes. the problem is just that i
don't really want to give up the inheritance design. is there a way
to maintain the inheritance that doesn't cause this huge
performance problem?

That is hard to say unless you post the rule and table schema you are
currently using.

Q <qdolan@gmail.com> wrote on 07/19/06 11:54 am:

On 19/07/2006, at 6:32 PM, Christian Rengstl wrote:

The analyze is from the exact query and i dropped the indexes
before the insert as well without imrpvement. The target table is
as well completely empty and the insert is supposed to write, in
this case, more or less 8 million lines in the table. There is a
rule though, because i have inheritance table structure with one
master table and around 20 child tables.

I would say the problem is in the rule. Try doing the insert into a
duplicate table with no rules or inheritance and see how long it
takes.

Perhaps you should provide the actual schema of tables and rules that
are involved in the query in question.

Q <qdolan@gmail.com> wrote on 07/19/06 4:37 am:

On 19/07/2006, at 4:24 AM, Christian Rengstl wrote:

now finally after a long time i have the query plan for the whole
filled table. I hope somebody can explain me why it takes so much
longer...

These explain analyze results don't appear to be from the
queries you
posted previously. For these results to mean anything you need to
include the EXACT queries you used to generate them.

Result (cost=0.06..254643.06 rows=8044000 width=39) (actual
time=0.056..655772
.273 rows=8044000 loops=1)
One-Time Filter: ((split_part(($1)::text, '_'::text,
2))::smallint = 1)
InitPlan
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.003..0.003 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.002..0.002 rows=1 loops=1)
-> Limit (cost=0.00..0.02 rows=1 width=10) (actual
time=0.006..0.007 rows
=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00
rows=8044000 width=
10) (actual time=0.004..0.004 rows=1 loops=1)
-> Seq Scan on temp_table (cost=0.00..194313.00 rows=8044000
width=39) (act
ual time=0.002..191672.344 rows=8044000 loops=1)
Total runtime: 62259544.896 ms

This is the query you want to be interested in, the others took no
time at all.

As a guess I would say the query is an INSERT INTO ... FROM
SELECT ... WHERE (split_part(???, '_', 2))::smallint = 1 statement.
The majority of the time appears to be taken up on the actual
INSERT
and not the SELECT part.

How many rows are in the target table and what indexes does it
have?
Does it have any triggers, check constraints, or rules applied
to it?
All these things can make the insert take longer as the number of
rows you have already in the table increases.

More than likely you have a target table with a LOT of rows and a
bunch of indexes on it and your disks are being thrashed because
the
indexes are not able to stay cached in RAM. At this point you
should
ensure your machine is not swapping do disk, and at the very least
you should go through one of the many tuning guidelines
available and
ensure you have allocated the appropriate amount of memory to
postgresql for your needs.

You may also want to consider dropping the indexes before you do
the
INSERT and recreate them afterwards.

"Christian Rengstl" <Christian.Rengstl@klinik.uni-
regensburg.de>
13.07.06 8.37 Uhr >>>

Good morning list,

the following query takes about 15 to 20 minutes for around 2
million lines in the file myfile.txt, but with 8 million lines it
takes around 5 hours and i just don't understand why there is such
a huge discrepancy in performance.

COPY public.temp_table FROM 'myfile.txt' WITH DELIMITER '\t';

INSERT INTO public.master(pid,smid, val1, val2, chr)
SELECT pid, smid, val1, val12, CAST(split_part((SELECT
chr from public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

INSERT INTO public.values(smid, pos, chr)
SELECT DISTINCT smid, pos, CAST(split_part((SELECT chr
from
public.temp_table LIMIT 1), '_', 2) as int2)
FROM public.temp_table;

I came up with this query, because i wanted to use the COPY
command
to load huge files into the db, but i don't want to load all the
columns contained in the file in only one table but copy some of
them into one table and some in a second table. As i found out
with
logging, the data is loaded into temp_table within 15 minutes, but
to transfer it from the temp_table toagain only something like 10
minutes. Can it be that the cast takes up so much more time than
when reading and transferring 2 million lines?

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan -
qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

---------------------------(end of
broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\

--
Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

On 19/07/2006, at 10:03 PM, Christian Rengstl wrote:

So here's the master table including the rules:

entry_no int8 NOT NULL DEFAULT nextval('public.master_seq'::regclass),
pid varchar(15) NOT NULL,
val_1 varchar(1),
val_2 varchar(1),
chr int2 NOT NULL,
aendat timestamp DEFAULT now(),
aennam varchar(8),
CONSTRAINT "PK_ENTRY" PRIMARY KEY (entry_no),
CONSTRAINT "UNIQUE_MASTER" UNIQUE (pid, entry_no)

CREATE OR REPLACE RULE "INSERT_INTO_1" AS
ON INSERT TO public.master
WHERE new.chr = 1 DO INSTEAD INSERT INTO public.table_1
(entry_no, pid, val_1, val_2, chr, aendat, aennam)
VALUES (new.entry_no, new.pid, new.val_1, new.val_2, new.chr,
new.aendat, new.aennam);

Like this i have around 20 rules so far, but there might be more
later on. The children tables are so far exactly as the master table.

What about the children? Do they have the same indexes?

You could try adding an 'ORDER BY chr' to your long running INSERT
INTO ... SELECT ... query.

Obviously it had something to do with the rule, because now
everything finished within 20 minutes. the problem is just that i
don't really want to give up the inheritance design. is there a way
to maintain the inheritance that doesn't cause this huge
performance problem?

When you say "now everything finished within 20 minutes", what did
you actually do to achieve this?

--
Seeya...Q

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

_____ / Quinton Dolan - qdolan@gmail.com
__ __/ / / __/ / /
/ __ / _/ / / Gold Coast, QLD, Australia
__/ __/ __/ ____/ / - / Ph: +61 419 729 806
_______ /
_\