More than once I have wished that mysqldump was never created. In many backup scenarios it is better to use other tools such as xtrabackup. However, there is still times when I end up using mysqldump. For one thing it is included with the mysql packages. It's "always there". Secondly, it sometimes is really a good fit for the tasks. And occassionally, with a bit of work, you can make it do neat tricks. 

I'm going to cover a few of the tricks I have used over the years. Hopefully you will find them helpful. In some cases one of these might even save your bacon.

Trick One:

One of the annoying issues with mysqldump is its lack of flexibility. It's very easy to dump all databases (--all-databases option), or one or more databases (--databases). Even dumping a single table isn't too difficult. But what about all the tables that start with "a"? You could specify each table on the command line, but I have ADD and I can't …

#!/bin/sh
# get.database.sizes.sh
# by dragkh
# Wed, 02 Mar 2011 17:59:44 +0100
# it will work on linux boxes with working mysql 
ps axu \
| grep datadir \
| grep mysql \
| grep -v grep \
| grep var \
| sed 's/^.*--datadir=//; s/ .*$//'   \
| sort \
| uniq  \
| while read crap
do 
echo ""
sized=$(du -s $crap | awk '{print $1}')
avail=$(df | grep $(dirname $crap) | awk '{print $2}')
echo -e  "$crap\t$sized\t$avail" | awk '{printf ("%s:\tUsed: %6.2fG \tDISK: %6.2fG\t Usage: %6.2f%% \n",$1,$2/1024/1024,$3/1024/1024,($2/$3)*100)}'
find $crap/ -maxdepth  1 -type d ! -type l  ! -path "$crap/"   -printf '"%p"\n'  | xargs  --no-run-if-empty  du -s | sort -nk 1,9 | awk -vavail=$avail '{crapy="";for(i=2; i<=NF; i++) {crapy=crapy" "$i;};printf (" -- %s\tUsed: %6.2fG \tDISK:%6.2fG Usage:%6.2f%%\n",crapy,$1/1024/1024,avail/1024/1024,($1/avail)*100)}'
done

result looks like

root@xxxx:[Wed Mar 02 16:54:59]:
/var/lib/mysql/aaaa/datafiles:               Used:  26.43G …

Restoring a single database from a full dump is pretty easy, using the mysql command line client’s --one-database option:

mysql> mysql -u root -p --one-database db_to_restore < fulldump.sql

But what if you don’t want to restore the database, you just want to extract it out of the dump file? Well, that happens to be easy as well, thanks to the magic of sed:

shell> sed -n '/^-- Current Database: `test`/,/^-- Current Database: `/p' fulldump.sql > test.sql

You just need to change “test” to be the name of the database you want extracted.