Sorry for such a basic question but what exactly is a Repeatable read as far
as transaction Isolation levels go. I understand the concept of serializable,
read committed & read uncommitted but i can't find any documentation that
describes the repeatable read level. How does this isolation level affect
concurrency. Sorry for such a stoopid question but it woud really help if I
got the low down once & for all. Many thanks in advance.
Ant
Repeatable read and serializable are very similar, and your question is far
from stupid.
In repeatable read, SQL Server will hold locks on the data you have read
until the end of the transaction. (Normally, in read committed, locks on
data you have read are released as soon as you are done reading.) So no one
can change change the data until you commit or rollback. Updates and deletes
are not permitted.
Serializable prevents INSERTS of new data that satisfy a condition that you
used for searching.
Example
BEGIN TRAN
SELECT * FROM titles
WHERE price < 5
-- suppose the above returns 2 rows; in repeatable read, those two rows are
now locked and no one can change them
-- until you commit or rollback; but someone in another connection could
insert a whole new row with price < 5
-- So, in serializable you are absolutely guaranteed that if you run the
exact SELECT again, you will get the same result:
SELECT * FROM titles
WHERE price < 5
-- But, in repeatable read, you might get more rows back, because of
inserts.
COMMIT TRAN
In repeatable read, SQL Server locks the data you read.
In serializable, SQL Server locks the data and the 'gaps' between the data.
HTH
Kalen Delaney
SQL Server MVP
www.SolidQualityLearning.com
"Ant" <Ant@.discussions.microsoft.com> wrote in message
news:78571FC8-E575-4D1C-BA59-E07C2C176A80@.microsoft.com...
> Sorry for such a basic question but what exactly is a Repeatable read as
> far
> as transaction Isolation levels go. I understand the concept of
> serializable,
> read committed & read uncommitted but i can't find any documentation that
> describes the repeatable read level. How does this isolation level affect
> concurrency. Sorry for such a stoopid question but it woud really help if
> I
> got the low down once & for all. Many thanks in advance.
> Ant
|||Kalen, thanks for your clear answer. I see the light! (I think)
So in fact, repeatable reads return more consistent data than read committed
as repeatable read locks the returning set until the transaction has either
been committed or rolled back,
:-
as opposed to read committed, which only locks the (committed) records
whilst the records are being read, therefore, the records in your recordset
may be updated by other transactions whilst it's sitting there in front of
you. correct?
Thanks for your help
Ant
|||Hi Ant
We're only talking about shared locks here for data that has been read (with
a SELECT). So there really isn't a concept of the rows themselves being
committed.
Rows that are modified acquire EXCLUSIVE locks, and exclusive locks are
always held until the end of the transaction, no matter what isolation level
you are in.
But your final statement is true. In read committed level, you can read a
record, and someone can change it while you decide what to do with it.
On the other hand, if you are in repeatable read or serializable level, you
are preventing all access to that row, and that can have a negative impact
on performance.
There are always tradeoffs.
HTH
Kalen Delaney
SQL Server MVP
www.SolidQualityLearning.com
"Ant" <Ant@.discussions.microsoft.com> wrote in message
news:7E7D398B-4FA9-4337-98CC-A0408FA2D494@.microsoft.com...
> Kalen, thanks for your clear answer. I see the light! (I think)
> So in fact, repeatable reads return more consistent data than read
> committed
> as repeatable read locks the returning set until the transaction has
> either
> been committed or rolled back,
> :-
> as opposed to read committed, which only locks the (committed) records
> whilst the records are being read, therefore, the records in your
> recordset
> may be updated by other transactions whilst it's sitting there in front of
> you. correct?
> Thanks for your help
> Ant
>
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment