Distribution log retation - log is huge!

Question

Hi,
I have 26 publishers with transactional replication to a central subscriber.
All the distributor properties and publication properties in these publishers are the same.
5 of these 26 publishers (publisher/distributor in the same server) have huge distribution log files (up to 100GB) and the log space used is over 90%, even using recovery model simple.
I already tried to configure the distribution retention and the publication retention, already checked the cleanup jobs and even executed the cleanup procedures myself, but the % space used in the huge logs are still over 90%.
I can understand a log retention in the publication database because the log reader agent needs to read the log and send the information to the distributor, but I can't understant this retention in the distributor database. As far I know, the information
in the distributor database is inside the system tables, the cleanup jobs delete old information in these tables, but why I still have this problem in the distribution log and only 5 of 26 distributors with this problem ?
How can I shrink the distribution log in this case ? The database is already in simple recovery model, I'm not understanding the log retention.
Thank you !
Dennes - Se resolveu, classifique a mensagem, por favor - [http://www.bufaloinfo.com.br] NOVO DVD Segurança no ASP.NET : http://www.bufaloinfo.com.br/LearingSeriesSegurancaASPNET2.asp

Answer 1

I had the same exact problem and tried everything thing: 
Problem:
Distribution data file = 19gb
Distribution log file = Ballooned to dangerous 700gb
Setup:
1) Transactional replication setup, continuous
2) 13 publications
3) Largest database = 200 gb
4) Distributor properties:
 Transaction retention 0-72 hours
 History Retention 48 hours
5) Subscriptions all configured as pull subscriptions
6) Distribution database is set to simple
7) Distribution clean up: distribution job run fine in 13 min(kinda long though)
8) Agent history clean up runs fine (2 sec)
9) Expired cleanup job runs fine (17 sec)
I noticed a few setup inneficiencies and recently made this change:
1) Changed each publication from "never expire" to  120 hour retention
Manually ran some diagnostics:
Segment Name Group Id Size in MB     Space Used Available Space Percent Used
distribution 1 19629.00       3635.81 15993.19  18.52
distribution_log 0 774523.06      741109.16 33413.91  95.69
DBCC OPENTRAN   
 No active open transactions
Manually ran EXEC dbo.sp_MSdistribution_cleanup @min_distretention = 0, @max_distretention = 72
Removed 0 replicated transactions consisting of 0 statements in 80 seconds (0 rows/sec).
DBCC Loginfo  
Returns 1104 rows
Select name, Log_reuse_wait_desc FROM sys.databasess
 Name             Log_reuse_wait_desc 
 distribution        LOG_Backup
 REPLuserdb        Replication
Manually confirmed all data from source to subscriber has been replicated and is a match. The data is all there at the subscriber.
I considered stopping the sql server agent and running the following command to reset the replication status: EXEC sp_repldone @xactid = NULL, @xact_segno = NULL, @numtrans = 0, @time= 0, @reset = 1    Then running the distribution cleanup
job again or shrinking the log.
....but this is very disruptive.
Finally I just changed the recovery model to full and back to simple and truncated the log file. Just like your case this is the only thing that worked!