Replication limitations over phone lines...?

[jballum]

Hi,

I was wondering if anybody could give me some information on Oracle's replication limitations when it comes to attempting to replicate between two DBS's that are not always connected to each other (ie. network connection down or phone line down).

Does Oracle have a built in 'queuing' process that will hold onto data that needs to be replicated and then when connection is obtained it will replicate at that time? Or is there a third party tool that is needed to handle this?

I have worked with Sybase before to handle this type of replication and they have separate products (can't remember what they are called at this time) that are needed that handles that data that needs to be replicated.

If you could point me to some link (or book) that I could read up on this (I have the documentation but it really doesn't get into this to well) it would be greatly appreciated.

Thanx,

JB

 

[lfoata]

If your network or modem link is down, Oracle will still write its archive files, and when the link is back up it will send the archives to the other server for synchronization

 

[Guest_imported]

Hi
Ifoata is talkin about standby databases.
U can also copy data using replication.
I think advanced replication is now part of the enterprise edition.
replication is based on triggers fireing at each commit for the tables U choose.
U have to read some manuals to find out what solution fits your environment.
regards
SteenB

 

[jballum]

Hi SteenB:

I think you misunderstood my question. I was asking for anybody's experiences when it came to replicating between two databases that are not always connected. The sorts of limitations they encountered or problems.

Like, once connections was finally obtained were any transactions lost. Or, after doing a lot of transactions did the queuing process Oracle uses to track transactions that were to be replicated fill up or fail. Things like that.

Any info on your experiences would be greatly appreciated.

JB

 

[uparna]

Hi jballum ,
I have a database that talks to another database that is situated kilometers away . As such , both the databases are not connected permanently and the databases talk to each other over the lines only when replication os going on . I am using Oracle8.1.7.1 version . I have faced some serious concerns , but none of them were serious . I will tell about these in detail .
The main problem with the replication going on is that the links are never stable. There are actualy three possibilities as fas as the link is concerned :
a. Link breaksdown before replication
b. Link breaks down during replication and
c. link breaks down after replication .

Option C offers no problem to us .

Problem in case a : Since i am doing complete refershes (I have complex snapshots ) When the replication is to happen when the link is down , then my database tries to access the Master database and on not finding it , simply truncates the data in these comples snapshots . I am given to believe that this does not happen in fast refreshes . In case this happens , we always revert back to our backups and keep the show going on . Once the link is up , I use the DBMS refresh facility to repopulate the data .

Problem in case b :One of the most dangerous scenarios. I have spent considerable time on trying to solve this ... as this causes serious damage . When the refreshes are going on and the link goes down , Oracle has behaved in two fashions - If the link goes down after a commit point is reached , then the replication simply stops and , half finished data will exist in the snapshot : for example one day i had found only 700 rows in a table having nearly a 5000 rows ! On the other side , if the link goes down when the replication is continuing and before a commit point is reached , the session just gets tristated ............. CAUSING SERIOUS CONCERN . When this happenned to me , any sort of a query on these snapshots , resulted in the session getting tristated ... the session just HUNG ! I couldnt even kill these sessions in any method : commands or using tools..... nothing works in a tristate ! Eventually i had to restart the database and everything started working perfectly . But let me also tell you that the probability of acse b occuring in the second fashion is quite low . In the past year , i had to just encountered it once........... not that i would like it occur again .

Please feel free to ask any thing else. WIll help you out to the best of my knowledge. Regards,
S. Jayaram Uparna .

If the need arises,you are welcome to mail me at oracguru@yahoo.com .

 

[jballum]

S. Jayaram Uparna ,

Thanks for all the info on the experiences/problems you have encountered. This is what I was looking for.

Have you had any experiences with multimaster replication with regards to the same problems you described above?

JB

 

[uparna]

Hi Jaballum ,
If you mean multiple databases by multimaster, then, unfortunately , i have not had any multimaster replication experience . Anyway , how does it change the scenario ? The only complication arises when there are more than one databases being contacted at the same time .. that would be tooo complicated . I am not sure weather a database can replicate with more than one database at a time . On the other hand (if Oracle talks to only one database at a time ) then the same cases that i had told you hold good.

If in case you mean multiple tables being accessed in a single snapshot (replica) then , let me tell you this : all my snapshots contact more than one tables .... approximately speaking , each one chooses data from more than 5 tables together .

Any thing more?


Regards,
S. Jayaram Uparna .