delete query hangs

ran a simple delete query that should have affected about 2.5 million = rows. The query was:
delete from traffic where left(filename, 4) =3D '0304'
After running this the TLOG increased to about 2 gig, but still had over = 5 gig of space left on the disk. Then the SPID for the query showed a = status of rollback this stayed like this for over 3 hours. After the = status changed to rollback I noticed one of my 4 processors was running = near 100% while the others were idle. I was forced to kill the query and = reboot the computer. After rebooting the same thing happened again.
Any ideas what is causing this?
I'm running SQL Server 2000There may have been an error somewhere, whether it was a deadlock, timeout,
out of space in the db(not set to grow) or whatever.
It's also possible that somebody killed the process.
Queries don't just roll back on their own.
Also, deleting 2.5 million rows can often be achieved by other means. That's
one big honkin' transaction.
Can you delete this data in a batch mode, say, 10,000 records at a time or
something?
You'd be amazed at the performance boost you can get from that technique
(although then you're not doing an atomic bit of work).
Another alternative method is to transfer the data from your source table
that _doesn't_ fit the criteria, and just drop the original table.
It all depends on your situation, though, such as if you've got steady data
flow into the table, or lots of updating, etc.
James Hokes
"jdoyle" <anonymous@.discussions.microsoft.com> wrote in message
news:03d101c3c5d2$212988c0$a601280a@.phx.gbl...
ran a simple delete query that should have affected about 2.5 million rows.
The query was:
delete from traffic where left(filename, 4) = '0304'
After running this the TLOG increased to about 2 gig, but still had over 5
gig of space left on the disk. Then the SPID for the query showed a status
of rollback this stayed like this for over 3 hours. After the status changed
to rollback I noticed one of my 4 processors was running near 100% while the
others were idle. I was forced to kill the query and reboot the computer.
After rebooting the same thing happened again.
Any ideas what is causing this?
I'm running SQL Server 2000|||I think James covered most in his post. Just one comment:
<<I was forced to kill the query and reboot the computer. >>
SQL Server was for some reason doing a rollback. And you now stop SQL Server, which mean that the
rollback will be restarted again at startup. If you think about it, SQL Server can't let a
transaction be partial committed... So, it could have been a good idea to let the on-going rollback
finish instead of restarting SQL Server. This is why MS added the WITH STATUSONLY option to the KILL
command in SQL Server 2000.
--
Tibor Karaszi, SQL Server MVP
Archive at: http://groups.google.com/groups?oi=djq&as_ugroup=microsoft.public.sqlserver
"jdoyle" <anonymous@.discussions.microsoft.com> wrote in message
news:03d101c3c5d2$212988c0$a601280a@.phx.gbl...
ran a simple delete query that should have affected about 2.5 million rows. The query was:
delete from traffic where left(filename, 4) = '0304'
After running this the TLOG increased to about 2 gig, but still had over 5 gig of space left on the
disk. Then the SPID for the query showed a status of rollback this stayed like this for over 3
hours. After the status changed to rollback I noticed one of my 4 processors was running near 100%
while the others were idle. I was forced to kill the query and reboot the computer. After rebooting
the same thing happened again.
Any ideas what is causing this?
I'm running SQL Server 2000|||I don't think the rollback was in progress. There appeared to be no disk =activity and one cpu just stayed at 100% usage for 3 hours. This query =should have taken about 15 minutes. Other applications such as EM =started to lock up during this time which is why I had to do the reboot. =
>--Original Message--
>I think James covered most in his post. Just one comment:
><<I was forced to kill the query and reboot the computer. >>
>SQL Server was for some reason doing a rollback. And you now stop SQL =Server, which mean that the
>rollback will be restarted again at startup. If you think about it, SQL =Server can't let a
>transaction be partial committed... So, it could have been a good idea =to let the on-going rollback
>finish instead of restarting SQL Server. This is why MS added the WITH =STATUSONLY option to the KILL
>command in SQL Server 2000.
>-- >Tibor Karaszi, SQL Server MVP
>Archive at: =http://groups.google.com/groups?oi=3Ddjq&as_ugroup=3Dmicrosoft.public.sql=
server
>
>"jdoyle" <anonymous@.discussions.microsoft.com> wrote in message
>news:03d101c3c5d2$212988c0$a601280a@.phx.gbl...
>ran a simple delete query that should have affected about 2.5 million =rows. The query was:
>delete from traffic where left(filename, 4) =3D '0304'
>
>After running this the TLOG increased to about 2 gig, but still had =over 5 gig of space left on the
>disk. Then the SPID for the query showed a status of rollback this =stayed like this for over 3
>hours. After the status changed to rollback I noticed one of my 4 =processors was running near 100%
>while the others were idle. I was forced to kill the query and reboot =the computer. After rebooting
>the same thing happened again.
>Any ideas what is causing this?
>I'm running SQL Server 2000
>
>.
>|||could you explain how to do the batch delete and the move method? Thanks
>--Original Message--
>There may have been an error somewhere, whether it was a deadlock, =timeout,
>out of space in the db(not set to grow) or whatever.
>It's also possible that somebody killed the process.
>Queries don't just roll back on their own.
>Also, deleting 2.5 million rows can often be achieved by other means. =That's
>one big honkin' transaction.
>Can you delete this data in a batch mode, say, 10,000 records at a time =or
>something?
>You'd be amazed at the performance boost you can get from that =technique
>(although then you're not doing an atomic bit of work).
>Another alternative method is to transfer the data from your source =table
>that _doesn't_ fit the criteria, and just drop the original table.
>It all depends on your situation, though, such as if you've got steady =data
>flow into the table, or lots of updating, etc.
>James Hokes
>
>"jdoyle" <anonymous@.discussions.microsoft.com> wrote in message
>news:03d101c3c5d2$212988c0$a601280a@.phx.gbl...
>ran a simple delete query that should have affected about 2.5 million =rows.
>The query was:
>delete from traffic where left(filename, 4) =3D '0304'
>
>After running this the TLOG increased to about 2 gig, but still had =over 5
>gig of space left on the disk. Then the SPID for the query showed a =status
>of rollback this stayed like this for over 3 hours. After the status =changed
>to rollback I noticed one of my 4 processors was running near 100% =while the
>others were idle. I was forced to kill the query and reboot the =computer.
>After rebooting the same thing happened again.
>Any ideas what is causing this?
>I'm running SQL Server 2000
>
>.
>|||What I do is like this:
/*Create a temporary table, say #_Keys*/
CREATE TABLE #_Keys
(
KeyValue int NOT NULL,
Ready bit NOT NULL DEFAULT(0),
Deleted bit NOT NULL DEFAULT(0)
)
/*Populate the temp table with the key values of the rows that you want to
delete*/
INSERT INTO #_Keys
SELECT
T.KeyValue,
0 AS Ready,
0 AS Deleted
FROM
MyTable AS T
WHERE
LEFT(T.filename, 4) = '0304'
CREATE INDEX SK_Ready ON #_Keys (Deleted, Ready) WITH FILLFACTOR=100
/*now you've got #_Keys populated and ready to go.*/
/*Next, start a WHILE loop like so:*/
WHILE EXISTS
(
SELECT
K.KeyValue
FROM
#_Keys AS K
WHERE
K.Deleted = 0
)
BEGIN
UPDATE
K
SET
K.Ready = 1
FROM
#_Keys AS K
INNER JOIN
(
SELECT
TOP 1000 K1.KeyValue
FROM
#_Keys AS K1
WHERE
K1.Ready = 0 AND
K1.Deleted = 0
) AS K2 ON
K2.KeyValue = K.KeyValue
/*Now, you've got 1000 rows marked and ready to delete.*/
/*Next, delete them*/
BEGIN TRANSACTION
DELETE
T
FROM
MyTable AS T
INNER JOIN #_Keys AS K ON
K.KeyValue = T.KeyValue
WHERE
K.Ready = 1 AND
K.Deleted = 0
UPDATE
K
SET
K.Deleted = 1
FROM
#_Keys AS K
WHERE
K.Ready = 1 AND
K.Deleted = 0
COMMIT TRANSACTION
END
DROP TABLE #_Keys
This will most likely delete the rows in less time than a normal DELETE
going after all 2.5 Million,
and it will likely cause less resource utilization for other concurrent
users as well.
Now, there are many other ways to do it; this is just one small example.
James Hokes
<anonymous@.discussions.microsoft.com> wrote in message
news:0a0301c3c641$fe3d7400$a501280a@.phx.gbl...
could you explain how to do the batch delete and the move method? Thanks
>--Original Message--
>There may have been an error somewhere, whether it was a deadlock, timeout,
>out of space in the db(not set to grow) or whatever.
>It's also possible that somebody killed the process.
>Queries don't just roll back on their own.
>Also, deleting 2.5 million rows can often be achieved by other means.
That's
>one big honkin' transaction.
>Can you delete this data in a batch mode, say, 10,000 records at a time or
>something?
>You'd be amazed at the performance boost you can get from that technique
>(although then you're not doing an atomic bit of work).
>Another alternative method is to transfer the data from your source table
>that _doesn't_ fit the criteria, and just drop the original table.
>It all depends on your situation, though, such as if you've got steady data
>flow into the table, or lots of updating, etc.
>James Hokes
>
>"jdoyle" <anonymous@.discussions.microsoft.com> wrote in message
>news:03d101c3c5d2$212988c0$a601280a@.phx.gbl...
>ran a simple delete query that should have affected about 2.5 million rows.
>The query was:
>delete from traffic where left(filename, 4) = '0304'
>
>After running this the TLOG increased to about 2 gig, but still had over 5
>gig of space left on the disk. Then the SPID for the query showed a status
>of rollback this stayed like this for over 3 hours. After the status
changed
>to rollback I noticed one of my 4 processors was running near 100% while
the
>others were idle. I was forced to kill the query and reboot the computer.
>After rebooting the same thing happened again.
>Any ideas what is causing this?
>I'm running SQL Server 2000
>
>.
>