Why is log file for tempdb filling up?

Question

Why is log file for tempdb filling up?

Bill Wagman

SSCommitted

Points: 1763
More actions
October 4, 2006 at 4:36 pm

#62440

Greetings,
In the application event log I am seeing the error 9002, Severity 17, State 6. The log file for the database tempdb is full. Back up the transaction log to free some space. I don't understand why this is happening. The tmepdb datafile is 3.6GB and is mostly unallocated. THere is 20GB free space on the drive on which it resides and the log file for the tempdb database is only 52MB. Both the datafile and the transaction log are set to autogrow by 10% and the file size is unlimited. So I don't understand why this is happening.
Poking around with google I find a couple of suggestions, one is that perhaps there is some query doing lots of sorting which could potentially cause the problem. Another suggestion has to do with timeing of backups and in fact the the regularly scheduled backup of the transaction log for the production database is started about 6 minutes before this event occurred. I also saw something about collating tables but I don't understand the relevance of that.
The question is twofold, 1) is there a method whereby I can determine what caused this and 2) what can I do to prevent it form occurring again?
Thanks.

Viewing 15 posts - 1 through 14 (of 14 total)

You must be logged in to reply to this topic. Login to reply

SQLBill SSC Guru Points: 51440 More actions · Answer 1

I do not recommend ever setting the growth to a percent. Set it to 100MB, see if that resolves the issue. If not, increase it to 200 or more.

There are many things that can use the TEMPDB. Usually that error just means the log file couldn't grow fast enough for the current activity.

-SQLBill

Chaohua Wang-372299 SSC Veteran Points: 209 More actions · Answer 2

This error may not cause your system down. The database tempdb is automatically shrunk normally. There is an "bug" here. We may get an error message if a user sends a large transaction while the database is shrinking. It happened.

Bill Wagman SSCommitted Points: 1763 More actions · Answer 3

SQLBill, I am curious as to why you suggest setting growth to a value rather than a percent. Are there preformance issues or is it merely because of the increasing uneven growth that makes possible?

Thanks.

SQLBill SSC Guru Points: 51440 More actions · Answer 4

Setting a percent will cause exponent growth. It's always growing 10% of the current size. So as the db gets larger it needs to expand even more.

Let's say you have a 10 mb db and you get 1 mb of growth a day. By 10%, the first growth will be 1 mb to 11 mb. The next growth will be 10% of that 1.1 mb to 12.1 mb, then 10% of that. Eventually, your db could be growing much larger than the space you really need. Let's say you have 10 GB of disk space and 9 gb is used. But you are auto-growing by 10%. The next autogrow will be 1 GB (remember you only need 1 mb a day) and now your database is full and cannot grow anymore. Adjusting it to grow by the proper MB size can help you to keep better control of the growth.

-SQLBill

Bill Wagman SSCommitted Points: 1763 More actions · Answer 5

Thanks, that is what I suspected.

I have seen many references to this problem being caused by bugs or large transactions or transactions occurring as the backup is running. Is there a means to prevent this problem or to at least determine what specific event caused it?

Thanks.

Sia King Shyan SSC-Addicted Points: 431 More actions · Answer 6

I have read that if autogrowth is set to a %, tempdb grows at 10% of the initial size. I always want to make sure this, but BOL doesn't explicitly specify it. Can anyone please confirm?

Thanks

SQLBill SSC Guru Points: 51440 More actions · Answer 7

Nope, everything I have seen and been told in MS Classes is that it grows by 10% of it's CURRENT size.

-SQLBill

SQLBill SSC Guru Points: 51440 More actions · Answer 8

From the BOL...

To specify that the file should grow by a percentage of the current file size, select By percent and specify a value.

-SQLBill

beadlesm SSC Enthusiast Points: 159 More actions · Answer 9

Tempdb issues in SQL 2000 : I do these two things for tempdb:

1) Apply MS KB recommendations for tempdb concurrency.[/url] This is done in 2005 as a production setup standard, too. To summarize, you create as many tempdb data files as there are physical processors, sized the same, and with autogrowth turned OFF. Log can be autogrow ON. You apply trace flag 1118 as a startup pramater.

For a 4 proc box, I usually set up 4 tempdb data files of 1 GB each. Log is set to 4 GB initial size with growth increments of 500 MB. I never set up autogrowth by %. Autogrow is really a failsafe thing - you don't want to depend on it to maintain your dbs. You should be sizing the dbs to allow for free space and monitor for low free space, to increase the free space after hours.

2) For tempdb that fills up fast -it is usually a developers job that is mis-using tempdb ... where a temp table is created and then one large transaction is launched, rather than a looped transaction with commits along the way.

Another cause I have seen is that replication (or a 3rd party replication product like Datamirror) using tempdb. Also in SQL 2005, certain index maint options can drive up tempdb to extremely large sizes very quickly.

In this case I put a SQL Agent job on the server to keep tempdb down -

step 1 in the job is simply to checkpoint tempdb.

step 2 is to truncate tempdb log.

Here is the script for the SQL 2000 job - try it in non-production first.[/b] I have been using it for two years in production on SQL 2000.

Adjust the schedule by observing tempdb in Enterprise Manager taskpad view and refresh the view every so often to see the data usage and log usage. In the case of this job script, tempdb log had grown to 135 GB, and the log fill rate was about a 500 MB every minute. I manually ran a checkpoint on tempdb, then shrank the log down to 4GB, then applied this job script to msdb db. I set the schedule for every 2 minutes, and so far tempdb has been fine.

(Haven't tried in SQL 2005 but should work fine).

[font="Arial Narrow"]

BEGIN TRANSACTION

DECLARE @JobID BINARY(16)

DECLARE @ReturnCode INT

SELECT @ReturnCode = 0

IF (SELECT COUNT(*) FROM msdb.dbo.syscategories WHERE name = N'Database Maintenance') < 1

EXECUTE msdb.dbo.sp_add_category @name = N'Database Maintenance'

-- Delete the job with the same name (if it exists)

SELECT @JobID = job_id

FROM msdb.dbo.sysjobs

WHERE (name = N'! Tempdb')

IF (@JobID IS NOT NULL)

BEGIN

-- Check if the job is a multi-server job

IF (EXISTS (SELECT *

FROM msdb.dbo.sysjobservers

WHERE (job_id = @JobID) AND (server_id <> 0)))

BEGIN

-- There is, so abort the script

RAISERROR (N'Unable to import job ''! Tempdb'' since there is already a multi-server job with this name.', 16, 1)

GOTO QuitWithRollback

END

ELSE

-- Delete the [local] job

EXECUTE msdb.dbo.sp_delete_job @job_name = N'! Tempdb'

SELECT @JobID = NULL

END

BEGIN

-- Add the job

EXECUTE @ReturnCode = msdb.dbo.sp_add_job @job_id = @JobID OUTPUT , @job_name = N'! Tempdb', @owner_login_name = N'sa', @description = N'MBEADLES: This job was placed on the server 1/8/2008 because tempdb Log file was growing very large (136 GB + ). This will checkpoint tempdb at 2 minute intervals and then truncate the tempdb log. I have used it on my prod servers for at least one year and no problems.', @category_name = N'Database Maintenance', @enabled = 1, @notify_level_email = 0, @notify_level_page = 0, @notify_level_netsend = 0, @notify_level_eventlog = 2, @delete_level= 0