DAO deadlock
shane -- shane@datatree.com
Friday, February 09, 1996
I need a little help here!
I have been using DAO with MS Access 2.0 and VC++ 4.0. I have been trying
to update records of the same database with three different applications.
Actually, they are all the exact same application and they are going through
the exact same sequence of events. The problem is that when they are trying
to update the same record, that record and/or page gets locked. I have set
up a try-catch sequence to catch the CDaoException with success. When I see
that I have caught an error, such as: "couldn't update, currently locked by
user "Admin" on computer ..." I try the update again, which is the solution
as explained by any of the "help" that I have found. The problem lies in
the fact that they all seem to be getting the exception with no one of them
actually having update access to the database. This throws me into an
infinite loop or deadlock. I have also set up a counter to show me how many
times it has tried to update the database. At some point in execution they
all continually grow.
bNotAccepted = TRUE
while (bNotAccepted)
{
try{
m_pHistorySet.Update();
bNotAccepted = FALSE;
}
catch (CDaoException* e) {
bNotAccepted = TRUE;
pEdit2->SetWindowText(e->m_pErrorInfo->m_strDescription);
e->Delete();
}
} /* end while */
Questions: Does anyone know a solution to this problem? Is it something
that I could be doing wrong or is it a "feature" of the DAO
design?
shane -- shane@datatree.com
Monday, February 12, 1996
Thanks HTH!
I should have mentioned also that I am opening the recordset as a dynaset(I
have also tried a tabletype). When the Update function for a recordset is
called when no BeginTrans or CommitTrans functions are used, the Update
function will commit all changes to the database automatically. Therefore,
each app is trying to commit its changes to the database but they can't as
no one of them are able to get update/write access to that record. Another
note: I have found my own workaround but am wondering why this needs to be
done and if there is a more efficient way of doing things.
My workaround was to: while using try/catch I would use a counter to see
how many times I had caught an error. After each error I would loop around
and try the Update again. After 25 attempts with an error I would throw it
out, close the recordset, reopen it again, and then try the update over with
success.
I don't know why I need to do this! It seems that I should have no problem
with just retrying the Update without closing the recordset. Why is it that
no one of the apps could get the access to the database to update it?
Appreciate the help!
>
>
>If I understand correctly what you've said, you have a situation where
>a database row is being updated simultaneously by multiple users. In
>this case you're absolutely correct - you're going to get a lock
>conflict. I believe that the only way to resolve the lock conflict is
>for each app to commit its changes to the database. HTH
>
>*************************** Attachment ***********************************
>Date: 11 Feburary 1996, 20:13:33 EST
>From: shane shane at INTERNET
> shane@datatree.com
>To: mfc-l at INTERNET
> mfc-l@netcom.com
>cc: frohnzie at INTERNET
> frohnzie@aol.com
>
>Reply-To: mfc-l at INTERNET
> mfc-l@netcom.com
>Subject: DAO deadlock
>X-Sender: shane
>Mime-Version: 1.0
>Content-Type: text/plain; charset="us-ascii"
>Precedence: list
>
>
>I need a little help here!
>
>I have been using DAO with MS Access 2.0 and VC++ 4.0. I have been trying
>to update records of the same database with three different applications.
>Actually, they are all the exact same application and they are going through
>the exact same sequence of events. The problem is that when they are trying
>to update the same record, that record and/or page gets locked. I have set
>up a try-catch sequence to catch the CDaoException with success. When I see
>that I have caught an error, such as: "couldn't update, currently locked by
>user "Admin" on computer ..." I try the update again, which is the solution
>as explained by any of the "help" that I have found. The problem lies in
>the fact that they all seem to be getting the exception with no one of them
>actually having update access to the database. This throws me into an
>infinite loop or deadlock. I have also set up a counter to show me how many
>times it has tried to update the database. At some point in execution they
>all continually grow.
>bNotAccepted = TRUE
>while (bNotAccepted)
>{
> try{
> m_pHistorySet.Update();
> bNotAccepted = FALSE;
> }
> catch (CDaoException* e) {
> bNotAccepted = TRUE;
> pEdit2->SetWindowText(e->m_pErrorInfo->m_strDescription);
> e->Delete();
> }
>} /* end while */
>
>
>Questions: Does anyone know a solution to this problem? Is it something
>that I could be doing wrong or is it a "feature" of the DAO
>design?
>
>
>
>
jarvisb@timken.com
Wednesday, February 14, 1996
I suspect that the reason none of your apps could successfully UPDATE
a record is because you tried to redo the UPDATE too quickly. Think
about it - if you detect a failure and immediately retry you're not
giving the system much time to get anything done. The time required
for the disk access is measured in multiples of a thousandth of a
second (milliseconds). The time required to execute the code is
measured in millionths of a second (microseconds), so you can probably
do MANY retries before the lock conflict might be resolved. A
reasonable thing to do is to set up a timer to expire after several
seconds, wait for that timer to expire, then retry your update. This
might work. My preferred solution when encountering a lock conflict
or deadlock is to roll back the transaction, wait for a few seconds,
and start over again. Hope this helps.
Bob Jarvis
*************************** Attachment ***********************************
Date: 14 Feburary 1996, 01:16:48 EST
From: shane shane at INTERNET
shane@datatree.com
To: mfc-l at INTERNET
mfc-l@netcom.com
Reply-To: mfc-l at INTERNET
mfc-l@netcom.com
Subject: Re: DAO deadlock
X-Sender: shane
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Precedence: list
Thanks HTH!
I should have mentioned also that I am opening the recordset as a dynaset(I
have also tried a tabletype). When the Update function for a recordset is
called when no BeginTrans or CommitTrans functions are used, the Update
function will commit all changes to the database automatically. Therefore,
each app is trying to commit its changes to the database but they can't as
no one of them are able to get update/write access to that record. Another
note: I have found my own workaround but am wondering why this needs to be
done and if there is a more efficient way of doing things.
My workaround was to: while using try/catch I would use a counter to see
how many times I had caught an error. After each error I would loop around
and try the Update again. After 25 attempts with an error I would throw it
out, close the recordset, reopen it again, and then try the update over with
success.
I don't know why I need to do this! It seems that I should have no problem
with just retrying the Update without closing the recordset. Why is it that
no one of the apps could get the access to the database to update it?
Appreciate the help!
>
>
>If I understand correctly what you've said, you have a situation where
>a database row is being updated simultaneously by multiple users. In
>this case you're absolutely correct - you're going to get a lock
>conflict. I believe that the only way to resolve the lock conflict is
>for each app to commit its changes to the database. HTH
>
>*************************** Attachment ***********************************
>Date: 11 Feburary 1996, 20:13:33 EST
>From: shane shane at INTERNET
> shane@datatree.com
>To: mfc-l at INTERNET
> mfc-l@netcom.com
>cc: frohnzie at INTERNET
> frohnzie@aol.com
>
>Reply-To: mfc-l at INTERNET
> mfc-l@netcom.com
>Subject: DAO deadlock
>X-Sender: shane
>Mime-Version: 1.0
>Content-Type: text/plain; charset="us-ascii"
>Precedence: list
>
>
>I need a little help here!
>
>I have been using DAO with MS Access 2.0 and VC++ 4.0. I have been trying
>to update records of the same database with three different applications.
>Actually, they are all the exact same application and they are going through
>the exact same sequence of events. The problem is that when they are trying
>to update the same record, that record and/or page gets locked. I have set
>up a try-catch sequence to catch the CDaoException with success. When I see
>that I have caught an error, such as: "couldn't update, currently locked by
>user "Admin" on computer ..." I try the update again, which is the solution
>as explained by any of the "help" that I have found. The problem lies in
>the fact that they all seem to be getting the exception with no one of them
>actually having update access to the database. This throws me into an
>infinite loop or deadlock. I have also set up a counter to show me how many
>times it has tried to update the database. At some point in execution they
>all continually grow.
>bNotAccepted = TRUE
>while (bNotAccepted)
>{
> try{
> m_pHistorySet.Update();
> bNotAccepted = FALSE;
> }
> catch (CDaoException* e) {
> bNotAccepted = TRUE;
> pEdit2->SetWindowText(e->m_pErrorInfo->m_strDescription);
> e->Delete();
> }
>} /* end while */
>
>
>Questions: Does anyone know a solution to this problem? Is it something
>that I could be doing wrong or is it a "feature" of the DAO
>design?
>
>
>
>
| Вернуться в корень Архива
|