The growth of Blue-Chase DBA (3): antique operation, data import and export become a problem

the growth of Blue-Chase DBA (3): antique operation, data import and export become a problem

Statement ***************************************

An individual's growth record on Oracle's path, with blue self-metaphor, to share growing emotions, horizons and technological changes and growth. Sensitive information is replaced in English, and will not disclose any corporate secrets, purely for technology sharing.

Creation is inspired by self-reflection and record. If you can have a little help or sympathy for the beginning of the library friends, gratified unceasingly.

Welcome to make bricks, such as the technical details of the error, please leave a message or e-mail ([email protected]) indicated, greatly appreciated.

***********************************************************************************

The meaning of life is not what you have now, but what you choose to pursue,

When a goal is pursued, life becomes full of meaning.

--Dark blue

Preface ***************************************

This is a personal record of the growth of the notes, since the step into the blue Sea of Oracle, inevitably all the way to the rush and constant test. By this story and friends to share the blue growth process.

I do not know when the blue has a kind of said not out of the obsession, obsessed with its breadth, obsessed with its depth, obsessed with the near and far away.

And don't know from when, looking at Oracle Red dazzling, illuminating a light, the unknown and confused at their feet began to show a little bit of life's enrichment and youth feedback.

Step forward on the path to the DBA's dream.

***********************************************************************************

August 4, 2014 Weihai

Contact the customer today, came to the data center A, the room environment is quite good. Unfortunately, when we see the actual machine, a little disappointed, no other ideas, just for its performance is more disappointing. The 09 machine, 5 years later, is arguably a few years old, but not too long. A server that was previously known to work for 20 years without downtime is still firm. And in front of these several, can only be said that the development of the IT sector is too fast, not enough, is not keeping up with the pace of the great times. And this time the server performance decline can be very simple explanation: in the initial stage, because of the pilot phase, the system daily business volume is very small, so for a long period of time have not burst out of the problem. And with the system of strong spread in this small branch of Weihai, the hidden trouble is still revealed. It is known that the server for nearly a year of small problems, and the field operators most commonly used method is: Restart the server! So I could not think of any place to replace, in order to use the word "antique" to express the mood at the moment. Next, let's look at the true colors of these servers:

(machine 1): Database server

Dell's Poweredge2950/e53201.86ghz Four thread/3.99g memory/136g/windows2003/oracle10g

(Machine 2): Application Server

Dell's Poweredge2950/e53201.86ghz Eight thread/7.99g memory/136g/windows2003

(Machine 3): FTP server

Dell's Poweredge2950/e53201.86ghz Four thread/3.99g memory/136g/windows2003

(Machine 4): storage

Dell's MD3000/1T/RAID1

Ridiculous configuration, compared to the highest performance of the server used to the application server, the database server and an FTP server the same configuration. It is strange to think that the "Cluster" logo, which is not torn off on two 4G memory machines, is like a set of clustered environments. Make sure that the two machines in front of you are really not clustered environments. With the engineers who are in charge of daily control, they get the answer. This is the plan to be removed, a third-party service provider has been there before, took over RAC, configured over storage. And later did not succeed to the withdrawal. Although got the answer, but see in front of the red warning lights on the flashing, or some noisy before what happened, really no words.

This scenario is inferred: the former third-party service personnel may be planning two machines to build a RAC cluster (Machine 1+ 3), using a slightly higher performance as an application server, storage using array storage, and security considerations, he did RAID1. But in reality due to the failure of the cluster construction, the adjustment of the scheme, into a single instance of the database, data files stored in the storage, and the application server is not changed, so that there is a server is idle, the final choice is to use it to build an FTP server for file sharing. According to this idea down, the scene of the present formation.

Think about it, if it is me: under Windows, the database cluster is not built, the performance of the server as a database server (machine 1), and then the remaining two servers to build an application cluster environment (machine 1+ machine 3), and, of course, the use of storage, the formation of a RAID 0 1, to improve performance bottlenecks when backing up data. To think, to do, but also landed on the needs of customers today.

Understand, the reason for the transformation platform is that the system is very unstable, performance reflected in the number of visitors to a long (50 people when the performance of the poor embodiment of the more obvious), there will be severe application response delay phenomenon.

After the leaders are present, the discussion meeting begins formally. After a discussion, the customer will be stability, safety as the most important consideration, the transformation plan can be decided by us, for the operator difficulty problem can be temporarily not considered. Here, I can foresee the one months of server downtime for the customer is what the suffering, as long as the stability, everything else can.

This time we propose two solutions:

Scenario One: The database server platform is transformed from Windows to Linux, and the application server still retains the Windows platform to reduce the difficulty of using the station operator;

Scenario Two: Database server platform, application server platform are changed to Linux platform to achieve stability, for operational difficulties, will provide guidance documents to the specific operators.

In the backup scenario, we propose two solutions:

Recommendation 1: Use Scheduled tasks to complete local automatic backup;

Recommendation 2: Remote backup disaster recovery, component last line of defense, and by our Beijing engineers to implement the backup.

After the confirmation, the transformation takes program two, the backup strategy combined with two recommendations, to achieve data backup disaster recovery of two lines of defense. Remote us as the last data security line of defense. In this regard, fought in the engine room.

Continuous use of similar viewing instructions to infer the output of the data: about 30G per year, the directive reference such as: "

SelectCount (1) from T where T.ziduan>=to_date (' 20090101 ', ' YYYYMMDD ') andt.ziduan<=to_date (' 20140101 ', ' YYYYMMDD '); "," Select sum (bytes)/1024/1024/1024from v$datafile; "," Select sum (bytes)/1024/1024/1024 from Dba_free_ Space; "

In the further decision to backup solution, see the Red warning light storage, really do not dare to use.

Try logical backup part of the data, storage read and write speed is quite slow, to determine the existence of storage problems. Looking at the output of the data, the original amount of data, we recommend that customers abandon the storage array, using the local server. Data growth can last for at least two years. Since the machine configuration is lower than the current one, it is recommended to upgrade the hardware system later. Customers are also adopted, the first to solve the system instability, upgrade plans are already considering the preparation. After getting the "call", the heart will be some end. Adoption Scenario: Database server (Machine 2), Application Server (machine 1), FTP component A local backup server (machine 3).

Now the question is: Back up the data files in the store, because it's really not practical to work on this dubious store. And it has been verified: initially wanted to dump the data directly from the store. 11:30 start export, 14:30 go back to see the amount of data exported 15G, to see the amount of data that needs to be exported 80G. It also means that all the data can be exported for more than 15 hours, which is a horrible thing. 80G of data is not big, so long, time is too long, so want to use other ways. Want to copy the data files on the system level to the local server, and then through similar instructions, such as: "ALTER DATABASE rename file ' z:/one.dbf ' to ' d:/one.dbf ';" Used to redirect data files and export files through a logical backup. This is exactly a cold spare, which needs to be done in the Guanqu situation.

In the process of copying data files to the server, it is also a painful process. Cold preparation starts from 16:00.

I do not know how much longer, cut off such as film:

Recall seems to be around 19:00, and finally finished, not expected fast, but at least more than directly by the storage guide a lot of practical. Now it's time to start working on the export.

Take a look at the time, start a logical backup, turn off the monitor, leave the return hotel and let the exported task finish at night. What can be done now is rest, May tonight, everything! ~ OK ~ ~.

*************************************** to be continued ***************************************

Welcome to: Blue Blog:http://blog.csdn.net/huangyanlong

************************************************** ***************************************