Viewing 15 posts - 76 through 90 (of 107 total)
1. create the transactional pub via SSMS GUI
2. create the transactional sub (with the same push/pull that you want) via SSMS GUI
3. generate scripts for same
4. edit and prettify scripts...
April 15, 2008 at 9:48 am
in SQL200x products replication is FREE (unlike competing vendors)
exceptions
1. if doing Oracle-SQL replication you would need SQL2005 Enterprise Edition
2. SQL/Express and SQLCE can be subscribers (not publisher or distribution engines)
BTW,...
April 15, 2008 at 9:44 am
can't check everything? must check everything? bug-free software? hmm !
forgot to answer Berto's last question - yes you can replicate from SQL2005 to SQL2000
there are some gotchas if trying to...
April 15, 2008 at 9:09 am
the LogReader (LR) reads from the AppDb tranlog (.LDF) and writes to the distribution db
if these App team changes are still going in, the supposition must be some blocking and/or...
April 15, 2008 at 9:04 am
Berto, you have said that there is a query that can max out your CPU for extended period. I am presuming this is in minority and that all other queries...
April 15, 2008 at 8:50 am
Barry raises good question ...
simplistically there are these processes on the Distributor
1. LogReader [reads from PUB.srcdb tranlog, and writes to DIST.distribution]
2. Distribution Agent (DA) [reads from distribution, and writes to...
April 14, 2008 at 11:57 am
1. local disks are fast
2. repl does row-at-once (lotsa round trips)
3. networks are slow
4. people send huge # of rows in one transactions
5. some subs are way over-indexed (but reporting...
April 14, 2008 at 11:42 am
1. using INT IDENTITY ... gives 2 ** 32 potential values (it's a lot!)
2. having long network latencies can mean that small allocations can be eaten up too quickly and...
April 14, 2008 at 11:31 am
I agree replication is wrong and backup(PROD)-restore(TEST) is correct, but consider ...
1. the disk geometry on PROD (data=E:, tlog=F:, dumps=G:) may differ from TEST (data=tlog=D:)
2. the maintplan on PROD probably...
April 14, 2008 at 11:21 am
you should investigate what the problem is
LogReader [reads from pub db and writes to distribution db]
Distribution Agent (DA) [reads from distribution and applies to sub db]
from what you have said,...
April 14, 2008 at 11:02 am
1. most PROD db's will be Recovery=FULL, probably with Log-Shipping to Disaster Recovery (D/R)
2. a 100% reload (500,000 x DELETE plus 500,000 INSERT) will be expensive
3. transactional repl (or...
April 14, 2008 at 10:05 am
1. as you do I/U/D statements a db will incur fragmentation [akin to breathing]
2. replication is chronological replay of what happened at pub on each sub
-> hence you...
April 14, 2008 at 9:43 am
replication works great apart from user fiddling !
left alone, the replication Distribution Agent (DA) plods along happily and replays I/U/D statements [via synthetic sprocs] and all is wonderful.
But then a...
April 14, 2008 at 9:21 am
the replication Distribution Agent (DA) has to faithfully "replay" at the sub whatever happened on the pub. This means it has to do a BEGIN TRAN ... stuff ... COMMIT...
April 14, 2008 at 9:12 am
1. it IS necessary to have a PK for transactional/merge replication
2. it is NOT necessary to have a PK for Snapshot repl
as Lowell rightly said, the Distribution Agent (DA) needs...
April 14, 2008 at 8:54 am
Viewing 15 posts - 76 through 90 (of 107 total)