Thiago
2011-01-31 17:09:44 UTC
Hello,
My production environment is running a SA 11.0.1.2527 database version
with HA option on Windows 2008R2 64-bits, where Primary and Mirror
servers are located in different cities (100Mbits link). There is just
one Database Server to controll three distinct databases, all of them
running in High Availability mode.
So far so good, the system is wonderful and the performance on commit
changes works fine (commits 10.000 rows in 0.5 seconds), but the
BACKUP is driving me crazy.
I´ve noticed that DBBACKUP -R does not work as I expected, so I´m
using the following statement:
BACKUP DATABASE DIRECTORY 'O:\\BACKUP' TRANSACTION LOG RENAME;
The backup is working ok, takes 4 minutes to backup a 7GB database.
The LOG RENAME statement proceeds as expected at the Mirror Server,
but the LOG file on the Primary Sever gets bigger after the RENAME
operation. In the first time I run the statement, the LOG file gets
around 40% bigger than original size. If I run it again, just after
the first time, the file shrinks as expected.
AFAIK, LOG RENAME should create a copy of the log file (YYMMDD%%.log)
and then start a new log file with the same original name. I can see
this happening, but in matter of 2 seconds the log file gets really
big again, only at the Primary Server, even when there is NO active
connections to the database... I´ve tested in development environment
and the behaviour is the same.
To test this abnormal behavior, I did a loop statement that insert
1.000.000 rows in a single table with a single integer column. The
file log grows around 20MB on both servers. After the first backup
statement, the log file gets around 30MB on Primary Server and 800KB
on Mirror Server.
For now, I´m running the BACKUP statement twice and I´m deleting all
renamed logs.
Any suggestions?!
Thank´s!
Thiago.
My production environment is running a SA 11.0.1.2527 database version
with HA option on Windows 2008R2 64-bits, where Primary and Mirror
servers are located in different cities (100Mbits link). There is just
one Database Server to controll three distinct databases, all of them
running in High Availability mode.
So far so good, the system is wonderful and the performance on commit
changes works fine (commits 10.000 rows in 0.5 seconds), but the
BACKUP is driving me crazy.
I´ve noticed that DBBACKUP -R does not work as I expected, so I´m
using the following statement:
BACKUP DATABASE DIRECTORY 'O:\\BACKUP' TRANSACTION LOG RENAME;
The backup is working ok, takes 4 minutes to backup a 7GB database.
The LOG RENAME statement proceeds as expected at the Mirror Server,
but the LOG file on the Primary Sever gets bigger after the RENAME
operation. In the first time I run the statement, the LOG file gets
around 40% bigger than original size. If I run it again, just after
the first time, the file shrinks as expected.
AFAIK, LOG RENAME should create a copy of the log file (YYMMDD%%.log)
and then start a new log file with the same original name. I can see
this happening, but in matter of 2 seconds the log file gets really
big again, only at the Primary Server, even when there is NO active
connections to the database... I´ve tested in development environment
and the behaviour is the same.
To test this abnormal behavior, I did a loop statement that insert
1.000.000 rows in a single table with a single integer column. The
file log grows around 20MB on both servers. After the first backup
statement, the log file gets around 30MB on Primary Server and 800KB
on Mirror Server.
For now, I´m running the BACKUP statement twice and I´m deleting all
renamed logs.
Any suggestions?!
Thank´s!
Thiago.