Oracle's ADG Needless to say that, in the words of the storage ring, are now stored from passive passive to active, but overall the software is being robbed, RAID was robbed ASM, snapshots are robbed Flashback, DR was robbed ADG.
If these types of programs that would combine what the outcome. This relates to a method of designing equipment library.
I also start a discussion. Environment are based on 11g of dg is.
First base, a main library a library equipment is used in many systems are in standby mode database design, if the database is more crucial, what are the disadvantages of this program yet.
11g of library equipment are now being given more responsibility.
Disaster recovery, disaster recovery if the first is the main library hang up standby database enables failover, this is no problem, then the standby database 11g now also been given more responsibility.
Batch query. There may be some batch query, reporting, query tasks. Then this part of the function can be placed by the library.
Prevention of human action. If there is a malfunction and human action, whether timely and reasonable precautions.
So the program will be a main one has several drawbacks,
First batch query, if the batch task pressure, large consumption of CPU resources for itself, if over many years, in fact, consumption is not negligible hardware itself, if the long run, may be a problem not necessarily cut over insurance.
If human errors, how to prevent this, if the delay too much afraid of losing data, do not delay, human error is inevitable, once a year, enough for us to take such a case the toss.
So we designed a two main backup?
If this design is also good for business-critical. However, there are several places to consider, or to meet the needs of just a few, and disaster recovery, and support inquiries, human error prevention.
So standby 1 can be designed in the same area adjacent to, but you can consider becoming standby2 site disaster recovery, remote disaster recovery is also very important to consider if the room down, several libraries are a preparation room , it is likely to be implicated.
So remote disaster recovery is also very important. So how to meet the needs of three above it.
First, disaster recovery, a primary two prepared, taking into account off-site disaster recovery, this in itself is a good solution.
Then support large queries, consider the people spread out a large query, run on two standby database, such as a 60%, a 40%, and so on, do not let the standby database completely idle. At the same time pressure also reduced a lot.
Then is a human error. How to prevent this, if the current one main two prepared environment, there are several options, one is a standby database open for another delay application archive.
For example, for off-site disaster recovery standby2, set the delay 2 hours, then appears in this case the human errors, such as truncate it is also possible to recover accidentally deleted data. But just maybe, because in the end a few hours no one can say. In case of misuse 3 hours ago, we finally found to be caused because of misuse, then this program is also too ball dropped.
Some friends will say, if within two hours of, such as switching the 20 archive, truncate operations are in the final archive, namely 20 archive operation, the data change section 20 archive It may resemble the following form.
truncate So how can you guarantee strict recovery back, some friends may say, can be used logminer, uh, you can, but still needs some work. And may not be able to 100% guarantee it.
What programs can continue to improve it. Try Flashback Database.
In fact, 11g, Flashback Database is not a new feature of. How he needs to complete the three above it.
Disaster recovery Needless to say, large queries can support, then it human error. That's below the green arrow.
On-site disaster recovery by 2 flashback to do, and then get the required data after crossing, then exp Export, and then deployed to the main library. Then the standby database applications can continue to open the log.
It seems this method was able to meet these needs.
So we then sort out what those needs again. So in this case, large queries can share about, and 1 on the standby database can share some more, on the standby database 2 can share some less. There is a need for off-site disaster recovery then add some additional space for the flashback logs.
So if the situation of the city room two machines Ben collapse, but more serious is that in a minute before the crash, there has been human operator accidentally deleted, then the off-site disaster recovery will not to force it.
In this case, you need to do first flashback database, exp data needs to come out, and then do failover, then turn off flashback database functionality, followed by redeployment to build disaster recovery environment, so I'd still be able to guarantee, there may be a little fly in the ointment is the possibility of flashback, if the table is relatively large, export the data may need to spend some time.
So the way a master prepared two can play so many tricks come out, as long as the stability of the program, so look Flashback Database in a primary two conditions prepared or could be considered.