1

Topic: Partial

Let's imagine basis on which the day data every day is flooded. Thus if the data  the certain period is there can be a week, month, year they do not participate in current processes in any way. Certainly, all of them are equally stored - for history, "just in case", etc.
And now we present that we want for UAT/SIT/DEV  to catch copies . Certainly "the old" data there is not necessary to us - they only take a place, extend process , etc. How to organize data storage and  to take only "the actual" data?
Well, for example, the most simple decision which comes to mind - for each table on  to create its copy ARCH_ the table in other , to organize automatic process of data migration in ARCH_ labels, and at  to specify to pass this .
The decision for certain works, but is too labor-consuming and is ugly.
Other method: we store in other  only partitions. Then copies of tables are not necessary, but I did not try and I do not know - whether the basis if at  to pass  opens UAT, and there are tables at which there are these partitions in it . Or all will be normal?
But for certain there are any other methods which to me simply do not come to mind - experience shows that many people can have set of judgements.

2

Re: Partial

Valergrad;
Transportable Tablespaces will be fastest probably, whether it is assured it is possible through Golden Gate to make it is made.
And what for on the test a basis bit? They will work with other data volume

3

Re: Partial

Valergrad wrote:

Other method: we store in other  only partitions. Then copies of tables are not necessary, but I did not try and I do not know - whether the basis if at  to pass  opens UAT, and there are tables at which there are these partitions in it . Or all will be normal?

DUPLICATE... SKIP FOREVER - opens with datafile offline drop (well or simply verifier )
But  it will not be possible to delete, sections of tables/indexes will not be remote yet
And it becomes through EXCHANGE PARTITION / DROP PARTITION
Then are deleted  and global indexes are rebuilt
Here only then announcements also begin that DBA  a grocery DB  - at me on DEV all flies

4

Re: Partial

Valergrad;
Can make on the contrary?
In PROD to store only actual given, and historical to transfer to the archive.
You early plan all:

wrote:

Well, for example, the most simple decision which comes to mind - for each table on  to create its copy ARCH_ the table in other , to organize automatic process of data migration in ARCH_ labels, and at  to specify to pass this .

It allows to discharge PROD.
p.s.
Partitions it is remarkable, but it is a paid option, it is bought from you?

5

Re: Partial

Vadim Lejnin wrote:

you early plan all:

Generally I just would like it to avoid it is simple an operation and hemorrhoids abyss smile

Vadim Lejnin wrote:

partitions it is remarkable, but it is a paid option, it is bought from you?

, certainly. Though from Lenin such suspiciousness is explainable smile

6

Re: Partial

Valergrad wrote:

it is passed...
Generally I just would like it to avoid it is simple an operation and hemorrhoids abyss smile
it is passed...
, certainly. Though from Lenin such suspiciousness is explainable smile

Partitions just also were invented what to avoid
Once adjusted and forgot
And then deduced irrelevant ts in offline, or
Or that it is more convenient

ALTER TABLE...
EXCHANGE PARTITION b...
WITH TABLE...
WITHOUT VALIDATION
UPDATE GLOBAL INDEXES;

7

Re: Partial

And how much it is a lot of data and what type? Not easier db_link  with DEV on PROM and result  with   in maidens?

8

Re: Partial

BTM;
Speech not, how something  and as on regular basis to create for UAT/SIT/DEV prod-like  - but without the superfluous data.

9

Re: Partial

Dr. SYS wrote:

Valergrad;
And what for on the test a basis bit? They will work with other data volume

Certainly, it is possible not to catch thus problem performance ( some subset performance of problems). For what, certainly to have a little  with the present copy  more correctly.
But - errors , errors in the code, damages of the data, to correlations, integration problems, etc. it is possible to catch all remaining problems and on "small" copies which owing to their size can be created on the regular basis and in a decent amount. A row performance of problems of type of curve requests or not created indexes will be visible and on them too.

10

Re: Partial

Valergrad;
Obviously, you did not understand, about what there is a speech in a sentence to use db_link. You can create on dev/uat/hz to basis procedure which using  from fighting basis through , will be regular  the data necessary to you in own tables. Has no value, from what basis you stretch _, once creating procedure and writing conditions , for example by date, you can always run it on any basis and obtain the necessary data in the necessary tables. A question only in what volume and type of this data.

11

Re: Partial

BTM, your method is not comprehensible on a large quantity of the reasons - select any!
1.  on  are forbidden in all serious organizations. Especially on maidens!
2. Volumes - ten terabyte, on a grid them you will not copy.
3. It will be difficult enough to write manually system which to trace - what data was already copied what is not present
Etc., etc. and so forth

12

Re: Partial

BTM wrote:

it is obvious, you did not understand

Obviously, you could not test this "approach" neither on the considerable volumes, nor on really loaded  servers.
I do not speak already about not so long ago considered purely logical problems in the approach.
2: besides decisions of class OGG and TTS still it is possible to do  an array on a stand, it was fulfilled somehow (successfully, by the way) by administrators such technology of cloning, but particulars I will not tell.
Still, probably, it is possible to try a little physical  to hold "under steam" - if the cooled down data on  to store strictly in Readonly tabular spaces, and on  which "under divorce" in time to do readonly-drop.
It was required sudden DEV - simply unhitched , renamed, opened and started up developers.
And on its place slowly tore the new smile
And if maidens-tests in a container DB to allocate as pluggable, how high boys in a court yard told, there the technology of cloning is for similar cases.

13

Re: Partial

Valergrad;
For the task to overtake one hundred gigabyte in another , db_link is quite comprehensible and even in the serious organizations them do from  on the test, observing ISSP, check. Seriously fresh data occupies tens from you terabyte? Anything difficult in a spelling  certainly is not present, but I do not insist on this approach certainly, you asked a method which was used by other people, here it, it the worker.
andrey_anonymous;
At all respect, be mistaken, it was possible to use such approach, but it is natural, not on tens terabyte.
We did not know initially an infrastructure and HARDWARE volumes.
At  volumes, the DB most likely is allocated on  class Netapp or EMC which  methods can  a DB for the test in some minutes. , by the way, good idea, in a combination with , including array means, it can appear an output at such data volumes.

14

Re: Partial

andrey_anonymous wrote:

And if maidens-tests in container DB to allocate as pluggable, how high boys in a court yard told, there the technology of cloning is for similar cases.

Well, at first, and  too thus it is necessary for convenience in container.
And secondly, which "" technology from for selection of a part of the data, different from the mentioned?

15

Re: Partial

AlexVin wrote:

which "" technology from for selection of a part of the data, different from the mentioned?

http://www.oracle.com/technetwork/datab … 078248.pdf
To honor
Clone PDB
Refreshable PDB
Still met any success story on Edition-based, but now near at hand the link is not present.

16

Re: Partial

andrey_anonymous wrote:

it is passed...
http://www.oracle.com/technetwork/datab … 078248.pdf
To honor
Clone PDB
Refreshable PDB
Still met any success story on Edition-based, but now near at hand the link is not present.

At us while 11.2 and passage will be very much at a distant day smile

17

Re: Partial

andrey_anonymous wrote:

it is passed...
To honor
Clone PDB
Refreshable PDB

Once again I will repeat - for selection of the PART of the data
Clone PDB - there it is possible as skip  to make
But to clone with such tables or their partitions, and others is not present - so it is impossible
By the way, for scenarios "clone-test-killed"  positioned

18

Re: Partial

AlexVin;
+Especial this clone pdb does not represent 1 anything. The same  a side view.  pdb like not bad. But if to count how many for it it will be necessary to pay - any desire to use disappears.
In general  and  our all for the maiden. If developers... Properly... To ask, it appears that is necessary to them at most percent of 2-5 % from all .
And for acceptance-loading tests -  with  forces  or .
It is necessary to divide simply correctly development and QA, not to hinder in one heap.

19

Re: Partial

AlexVin wrote:

it is passed...
Once again I will repeat - for selection of the PART of the data
Clone PDB - there it is possible as skip  to make
...
By the way, for scenarios "clone-test-killed"  positioned

Not the reader... smile

20

Re: Partial

- stick with a nose

21

Re: Partial

Valergrad;
bases means of the watchman as already told, rescue the father of democracy. Turn out almost quickly and places do not occupy almost. All  so did, on Natapp-e, for example. Well and Orakel, in my opinion, is able . But how many it will cost? Did not check.

22

Re: Partial

23

Re: Partial

24

Re: Partial

andrey_anonymous wrote:

Developers and testers mold to themselves "shareware" (on resources) clones on need

It is all it is thin. At us it is licensed and QA in force http://www.oracle.com/us/corporate/pric … 070584.pdf
this limited license allows the user to develop applications using the licensed products as long as such applications have notbeen used for any data processing, business, commercial, or production purposes
It is obvious, what the clone  just gets under it, differently what for it is necessary?