 |
IDS Forum
Re: Question on HPL --- FAILURE IN LOADING TO ....
Posted By: Art Kagel
Date: Friday, 19 October 2012, at 10:58 a.m.
In Response To: Re: Question on HPL --- FAILURE IN LOADING TO TARG (Kern Doe)
Two informix instances, even on the same machine, cannot talk to each other
through shared memory connections. You MUST use a network or streams
connection to the 'other' instance. You can connect to the instance where
you are running the SQL using shared memory, but the remote server alias
that you use in the remote insert (or remote select) must be a non-shared
memory connection.
BTW, if you are copying data from a table in one instance into a table in
another instance, use my dbcopy utility. It is about 2X as fast as INSERT
INTO ... SELECT ... FROM when the tables reside in different instances
(even faster if the other instance is on another machine). Dbcopy also
avoids the long transaction problem if you are loading into a 'standard'
table (as opposed to RAW). Also, if you can break the ranges of keys down,
you can run several copies of dbcopy, each copying a distinct range of
keys, then you can multiply the speed gain almost linearly until you run
out of resources (become IO or CPU bound). Normally I will copy a big
table with from four to ten copies of dbcopy running (I use the dbschema
-hd output to select ranges of keys that are roughly equal in size and pass
appropriate WHERE clauses or full SELECTs to each copy).
Copying between instances on separate machines, running dbcopy on a third
machine to get access to additional CPU resources, I've run up to 40 copies
of dbcopy in parallel on a 32 core machine - WOOSH!
Dbcopy is included in the package utils2_ak in the IIUG Software Repository.
Art
Art S. Kagel
Advanced DataTools (www.advancedatatools.com)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of
other individuals affiliated with any entity with which I am affiliated nor
those of the entities themselves.
On Fri, Oct 19, 2012 at 10:43 AM, Kern Doe <kern_doe@yahoo.com> wrote:
> Art,
> Your detail notes and experience on RAW TABLE to share are certainly
> appreciated.
> My original topic was "HPL", has become an experiment on raw, and then now
> I
> have to ask a question on unrelated or somewhat related topic.
> OK, so yesterday I tried to load those 250 mil rows (from an external
> table)
> into the target table of different instance but in the same host, and I ran
> into error 908, I even tried tcp protocol (which I should not since both
> instances are in the same box), it crashed the target instance. Any idea?
>
> SQL: New Run Modify Use-editor Output Choose Save Info Drop Exit
> Modify the current SQL statements using the SQL editor.
> ----------------------- target_db@target_instance --------- Press CTRL-W
> for
> Help --------
> INSERT INTO target_table SELECT * FROM
> source_db@source_instance:external_source_table
>
> 908: Attempt to connect to database server (source_instance) failed.
>
> ________________________________
> From: Art Kagel <art.kagel@gmail.com>
> To: ids@iiug.org
> Sent: Thursday, October 18, 2012 8:19 PM
> Subject: Re: Question on HPL [28573]
>
> Kern:
>
> Very important to keep in mind when working with RAW tables:
>
> 1. RAW mode tables do not support PRIMARY or UNIQUE key constraints.
>
> 1. You will have to drop those and recreate them after altering the
>
> table's mode back to "STANDARD".
>
> 2. If those constraints rely on "hidden" indexes (ones whose names
>
> begin with a space) the indexes will have to be rebuild also so this may
>
> add runtime.
>
> 3. If there are foreign keys on other tables that reference the
>
> dropped constraints, those foreign key constraints will drop as well and
>
> will have to be recreated which may mean rebuilding the
> supporting indexes
>
> on those keys (if the indexes are hidden) and will require that
> all of the
>
> keys in the dependent table be re-validated against the table you were
>
> loading.
>
> 4. This means that the total runtime of a load may actually be longer
>
> if you alter it to "RAW" and back to "STANDARD" before you can put the
>
> database back into service.
>
> 2. If you load into a table in "RAW" mode, you have to take a level
>
> zero archive after altering the table back to "STANDARD" mode so that you
>
> can recover after a hard crash without having to repeat the data load.
>
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinions
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those of
> other individuals affiliated with any entity with which I am affiliated nor
> those of the entities themselves.
>
> On Thu, Oct 18, 2012 at 8:58 PM, Kern Doe <kern_doe@yahoo.com> wrote:
>
> > I've tried this unloading involved "raw table" -- the result has blown my
> > mind
> > away.
> > My first test result was 9 minutes to unload 63 mil rows while hpl took
> > about
> > 1 hour.
> > My 2nd test was 38 min for 250mil rows, amazing!! I will test the load
> > soon.
> >
> > Reinhard, thank you for the script, I haven't tried it, there are some
> > parts I
> > believe you meant differently, please see my comments.
> > Thank you Reinhard.
> >
> > ________________________________
> > From: "Habichtsberg, Reinhard" <RHabichtsberg@arz-emmendingen.de>
> > To: ids@iiug.org
> > Sent: Thursday, October 18, 2012 2:47 AM
> > Subject: RE: Question on HPL [28562]
> >
> > If the tables resp. the unload files will grow too big you may even
> > unload and load using external tables and pipes with no need of storage
> > space:
> >
> > Solaris 10, Informix 11.70.FC5
> > Example shell script:
> > #!/usr/bin/bash
> > # Target table was created in target database and has same schema as
> > source table
> > # Two Informix Server: source_server and target_server, may also work if
> > source_server and target_server and source_database and target_database
> > are the same
> > # FORMAT 'informix' may not work for tables with byte or text columns
> >
> > # Target table in target database set to type raw due to express mode:
> > dbaccess target_database@target_server - << EOF
> > alter table target_table type (raw);
> > EOF
> >
> > # Create pipe for loading:
> > mkfifo /tmp/target_table.lo
> >
> > # Create pipe for unloading:
> > mkfifo /tmp/target_table.un
> >
> > # Create external table in target database:
> > dbaccess target_database@target_server - << EOF
> > CREATE EXTERNAL TABLE ext_target_table sameas target_table
> > USING (DATAFILES ('PIPE:/tmp/target_table.lo'),
> > FORMAT 'informix',
> > REJECTFILE 'target_database_target_table.log',
> > EXPRESS);
> > EOF
> >
> > # Create external table in source database:
> > dbaccess source_database@source_server - << EOF
> > CREATE EXTERNAL TABLE ext_source_table SAMEAS source_table
> > USING (DATAFILES ('PIPE:/tmp/source_table.un'),
> > FORMAT 'informix');
> > EOF
> >
> > # Open pipe in background:
> > cat > /tmp/$tabelle.lo < /tmp/$tabelle.un &
> >
> > (you probably meant cat > /tmp/target_table.lo < /tmp/target_table.un &)
> >
> > # Insert Import Command in background:
> > dbaccess target_database@target_Server - <<EOF &
> > insert into target_table select * from ext_target_table;
> > EOF
> >
> > # Sleep 3 seconds
> > sleep 3
> >
> > # Insert Export Command:
> > dbaccess source_database@source_server - <<EOF
> > insert into ext_source_table select * from source_table;
> > EOF
> >
> > echo EXPORT/IMPORT done!
> > # Sleep 5 seconds
> > sleep 5
> >
> > # Clean up
> > dbaccess source_database@source_server - <<EOF
> > drop table ext_source_table;
> > EOF
> > dbaccess target_database@target_Server - <<EOF
> > drop table ext_target_table;
> > EOF
> >
> > echo "drop PIPE:"
> > rm /tmp/ target_table.lo
> > rm /tmp/source_table.un
> >
> > # Set Target Table to type standard
> > dbaccess target_database@target_Server - <<EOF
> > alter table $tabelle type (standard); (you meant target_table?)
> > EOF
> >
> > I scipped the error handling in due of clarity.
> >
> > HTH
> > Reinhard
> >
> > > -----Original Message-----
> > > From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of
> > Art
> > > Kagel
> > > Sent: Thursday, October 18, 2012 2:25 AM
> > > To: ids@iiug.org
> > > Subject: Re: Question on HPL [28556]
> > >
> > > You can load using external tables also. Say you have a delimited
> > file, mydata.unl,
> > > that you want to load into a table named mydata. So you define an
> > external table
> > > defined over the file mydata.unl and insert into the normal table
> > select from the
> > > external table. So:
> > >
> > > create external table mydata_unl sameas mydata using (datafiles(
> > > "disk:/some/path/mydata.unl"), format "delimited", delimiter "|" );
> > >
> > > insert into mydata
> > > select * from mydata_unl;
> > >
> > > Art
> > >
> > > Art S. Kagel
> > > Advanced DataTools (http://www.advancedatatools.com/)
> > > Blog: http://informix-myview.blogspot.com/
> > >
> > > Disclaimer: Please keep in mind that my own opinions are my own
> > opinions and
> > > do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> > other
> > > organization with which I am associated either explicitly, implicitly,
> > or by
> > > inference. Neither do those opinions reflect those of other
> > individuals affiliated with
> > > any entity with which I am affiliated nor those of the entities
> > themselves.
> > >
> > > On Wed, Oct 17, 2012 at 6:51 PM, Kern Doe <kern_doe@yahoo.com> wrote:
> > >
> > > > This is great! How would you guys load?
> > > > Would you use dbload or HPL, I know I don't want to use "load from".
> > > >
> > > > Any way, thank you John/Art, we'll try this tomorrow.
> > > > Kern --
> > > >
> > > > ________________________________
> > > > From: John Miller iii <miller3@us.ibm.com>
> > > > To: ids@iiug.org
> > > > Sent: Wednesday, October 17, 2012 5:09 PM
> > > > Subject: Re: Question on HPL [28552]
> > > >
> > > > Here is the syntax for unloading the customer table to 2 files.
> > > >
> > > > CREATE EXTERNAL TABLE 'informix'.EXTcustomer SAMEAS
> > > > 'informix'.customer USING (
> > > > DATAFILES('DISK:/tmp/customer_1.unl','DISK:/tmp/customer_2.unl'),
> > > > FORMAT 'DELIMITED',
> > > > DELIMITER '|',
> > > > RECORDEND '',
> > > > ESCAPE
> > > > );
> > > >
> > > > INSERT INTO 'informix'.customer
> > > > SELECT * FROM 'informix'.EXTcustomer;
> > > >
> > > > John F. Miller III
> > > > STSM, Embedability Architect
> > > > miller3@us.ibm.com
> > > > 503-578-5645
> > > > IBM Informix Dynamic Server (IDS)
> > > >
> > > > ids-bounces@iiug.org wrote on 10/17/2012 11:54:06 AM:
> > > >
> > > > > From: "Art Kagel" <art.kagel@gmail.com>
> > > > > To: ids@iiug.org,
> > > > > Date: 10/17/2012 11:55 AM
> > > > > Subject: Re: Question on HPL [28551] Sent by: ids-bounces@iiug.org
> > > > >
> > > > > Note, this is typed off the cuff, so I may have missed a quote or
> > > > > comma somewhere.
> > > > >
> > > > > create external table sometable_ext like sometable using
> > > > > (datafiles("disk:/some/path/file","disk:/other/path/file"));
> > > > >
> > > > > select * from sometable insert into sometable_ext;
> > > > >
> > > > > VOILA!
> > > > >
> > > > > For import, when you create the external table you can add the
> > > > > EXPRESS
> > > > mode
> > > > > attribute and it will load the data the same way that HPL in
> > express
> > > > > mode
> > > >
> > > > > does. You can change the data format from DELIMITED, to INFORMIX,
> > or
> > > > > FIXED. You can change the delimiter for the default delimited
> > format
> > > > > - no
> > > >
> > > > > you can't get rid of the trailing delimiter 8-( - etc. You can
> > write
> > > > > to a
> > > >
> > > > > pipe so the data can be compressed. Check out the description of
> > > > > CREATE EXTERNAL TABLE in the online or PDF manuals for full
> > details.
> > > > >
> > > > > Art
> > > > >
> > > > > Art S. Kagel
> > > > > Advanced DataTools (http://www.advancedatatools.com/)
> > > > > Blog: http://informix-myview.blogspot.com/
> > > > >
> > > > > Disclaimer: Please keep in mind that my own opinions are my own
> > > > > opinions and do not reflect on my employer, Advanced DataTools,
> > the
> > > > > IIUG, nor any other organization with which I am associated either
> > > > > explicitly, implicitly, or by inference. Neither do those opinions
> > > > > reflect those of other individuals affiliated with any entity with
> > > > > which I am affiliated
> > > > nor
> > > > > those of the entities themselves.
> > > > >
> > > > > On Wed, Oct 17, 2012 at 2:44 PM, Kern Doe <kern_doe@yahoo.com>
> > wrote:
> > > > >
> > > > > > This is getting more interesting. Do you have a simple sample
> > > > > > codes involved unload/load with external tables? I tried HPL
> > > > > > unload yesterday on 63
> > > > mil
> > > > > > rows
> > > > > > tables, it took 1hr15min and already happy about it, anything
> > else
> > > > > > I
> > > > can
> > > > > > do to
> > > > > > improve further the speed and the flexibility, I'm all for it.
> > > > > > Thanks Art.
> > > > > >
> > > > > > ________________________________
> > > > > > From: Art Kagel <art.kagel@gmail.com>
> > > > > > To: ids@iiug.org
> > > > > > Sent: Wednesday, October 17, 2012 1:19 PM
> > > > > > Subject: Re: Question on HPL [28546]
> > > > > >
> > > > > > FYI, if you have 11.50 or later, you can use external tables for
> > > > > > the
> > > > unload
> > > > > > and it is at least as fast as HPL. Then the select to filter the
> > > > > > data being written into the external table would be trivial.
> > > > > >
> > > > > > Art
> > > > > >
> > > > > > Art S. Kagel
> > > > > > Advanced DataTools (http://www.advancedatatools.com/)
> > > > > > Blog: http://informix-myview.blogspot.com/
> > > > > >
> > > > > > Disclaimer: Please keep in mind that my own opinions are my own
> > > > opinions
> > > > > > and do not reflect on my employer, Advanced DataTools, the IIUG,
> > > > > > nor
> > > > any
> > > > > > other organization with which I am associated either explicitly,
> > > > > > implicitly, or by inference. Neither do those opinions reflect
> > > > > > those of
> > > >
> > > > > > other individuals affiliated with any entity with which I am
> > > > > > affiliated
> > > > nor
> > > > > > those of the entities themselves.
> > > > > >
> > > > > > On Wed, Oct 17, 2012 at 1:32 PM, Kern Doe <kern_doe@yahoo.com>
> > > wrote:
> > > > > >
> > > > > > > Dear HPL experts,
> > > > > > > I create a simple HPL unload, it works fast, and it basically
> > > > > > > dumps
> > > > the
> > > > > > > entire
> > > > > > > table. Is there way I can customize HPL unload for selected
> > > > > > > data, for
> > > >
> > > > > > > example,
> > > > > > > "customer_id < 2000000" ?
> > > > > > > Thanks in advance.
> > > > > > > Kern --
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > >
> > > >
> > > >
> > > >
> > >
> > ************************************************************************
> > *******
> > > >
> > > > > > > Forum Note: Use "Reply" to post a response in the discussion
> > forum.
> > > > > > >
> > > > > > >
> > > > > >
> > > > > > --e89a8f923ba0ddf2be04cc454df1
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > >
> > > >
> > > >
> > > >
> > >
> > ************************************************************************
> > *******
> > > >
> > > > > > Forum Note: Use "Reply" to post a response in the discussion
> > forum.
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > >
> > > >
> > > >
> > > >
> > >
> > ************************************************************************
> > *******
> > > >
> > > > > > Forum Note: Use "Reply" to post a response in the discussion
> > forum.
> > > > > >
> > > > > >
> > > > >
> > > > > --14dae93404b93508a004cc45cbb6
> > > > >
> > > > >
> > > > >
> > > >
> > > >
> > > >
> > > >
> > >
> > ************************************************************************
> > *******
> > > >
> > > > > Forum Note: Use "Reply" to post a response in the discussion
> > forum.
> > > > >
> > > >
> > > >
> > > >
> > > >
> > >
> > ************************************************************************
> > *******
> > > > Forum Note: Use "Reply" to post a response in the discussion forum.
> > > >
> > > >
> > > >
> > > >
> > >
> > ************************************************************************
> > *******
> > > > Forum Note: Use "Reply" to post a response in the discussion forum.
> > > >
> > > >
> > >
> > > --90e6ba614c8042c2e504cc4a69e1
> > >
> > >
> > >
> > ************************************************************************
> > *******
> > > Forum Note: Use "Reply" to post a response in the discussion forum.
> >
> >
> >
> >
>
>
> *******************************************************************************
> > Forum Note: Use "Reply" to post a response in the discussion forum.
> >
> >
> >
> >
>
>
> *******************************************************************************
> > Forum Note: Use "Reply" to post a response in the discussion forum.
> >
> >
>
> --047d7b15a57dca84da04cc5f4b7d
>
>
>
> *******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
>
> *******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--e89a8f838b35d0bf7404cc6abc89
Messages In This Thread
IDS Forum is maintained by Administrator with WebBBS 5.12.
|
|
