Installation fails, possible mysql problem

[webminster]

Installing on Fedora Core 3... Get to the install step "Starting Servers..." and install hangs. I note in the mysql log the following:

InnoDB: Setting log file ./ib_logfile0 size to 14 MB
InnoDB: Database physically writes the file full: wait...
InnoDB: Error: ib_logfiles are too small for innodb_thread_concurrency 200.
InnoDB: The combined size of ib_logfiles should be bigger than
InnoDB: 200 kB * innodb_thread_concurrency.
InnoDB: To get mysqld to start up, set innodb_thread_concurrency in my.cnf
InnoDB: to a lower value, for example, to 8. After an ERROR-FREE shutdown
InnoDB: of mysqld you can adjust the size of ib_logfiles, as explained in
InnoDB: http://dev.mysql.com/doc/mysql/en/Ad..._removing.html
InnoDB: Cannot continue operation. Calling exit(1).
050908 20:59:52 mysqld ended

Any way to get past this? Thanks.
-Alan

[KevinH]

Hi,

Can you run df -k and check your available disk space?

thanks

-kevin

[webminster]

[asparks@phoenix ~]$ df
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/hda2 20161204 15959068 3177996 84% /
/dev/hda1 101086 22552 73315 24% /boot
none 94832 0 94832 0% /dev/shm
/dev/shm 153600 28 153572 1% /opt/zimbra/amavisd/tmp

Thanks.

[marcmac]

Originally Posted by webminster

Installing on Fedora Core 3... Get to the install step "Starting Servers..." and install hangs. I note in the mysql log the following:

InnoDB: Setting log file ./ib_logfile0 size to 14 MB
InnoDB: Database physically writes the file full: wait...
InnoDB: Error: ib_logfiles are too small for innodb_thread_concurrency 200.
InnoDB: The combined size of ib_logfiles should be bigger than
InnoDB: 200 kB * innodb_thread_concurrency.
InnoDB: To get mysqld to start up, set innodb_thread_concurrency in my.cnf
InnoDB: to a lower value, for example, to 8. After an ERROR-FREE shutdown
InnoDB: of mysqld you can adjust the size of ib_logfiles, as explained in
InnoDB: http://dev.mysql.com/doc/mysql/en/Ad..._removing.html
InnoDB: Cannot continue operation. Calling exit(1).
050908 20:59:52 mysqld ended

Any way to get past this? Thanks.
-Alan

It looks like you don't have enough memory - by default, we assign 40% of available system memory to mysql. The log files are sized at 20% of that value - so, if you've got limited RAM, the log file sizes are too low for the concurrency value.

To fix this, edit /opt/zimbra/conf/my.cnf and lower innodb_thread_concurrency from 200. THe error message shows that your logfiles are 14M, so concurrency*200kB should be less than that - so a value of 50 should work.

Additionally, you'll want to lower max_connections in my.cnf to a smaller value - 50 should work for that, too.

This is being tracked in bug 3911

[webminster]

How do I do this during running install.sh? How do I pre-change the configuration so it'll start the DB properly during the "initializing db" phase? The installation is corrupt since the database won't come up for initialization...
-Alan

[marcmac]

What you need to do is modify the innodb_thread_concurrency and max_connections prior to mysql installation.

You'll need to modify the install.sh script.

Before the line that runs "zmmyinit" (line 989, I think) you'll want to add these two lines (inside the conditional block):

runAsZimbra "zmlocalconfig -e mysql_innodb_thread_concurrency=50"
runAsZimbra "zmlocalconfig -e mysql_max_connections=50"

This will set the parameters prior to database creation.

Originally Posted by webminster

How do I do this during running install.sh? How do I pre-change the configuration so it'll start the DB properly during the "initializing db" phase? The installation is corrupt since the database won't come up for initialization...
-Alan

[webminster]

Thanks! Good call, that worked nicely.
-Alan