Why do you need this?

Just suppose you want to backup a Mongo Database using mongodump.

As backups for individual databases don’t support –oplog, you may want to make sure you get a consistent database and therefore you decide to use fsyncLock() to flush any pending write operation and to lock new writes until your task is finished (do this only over secondary nodes on replica sets)

But then you launch mongodump against the MongoDB node… and the task keeps waiting for something…

Guess what? Most probably the database you want to backup doesn’t exist.  But as mongodump will try to extract data, this will trigger the database autocreation.

Look to the following example:

> show dbs > use testdb switched to db testdb > show dbs > db.testcol.find() > show dbs testdb (empty)

As you used fsyncLock(), mongodump will keep waiting until it can create the new database. But that never will happen unless you run fsyncUnlock() by hand.

So the key is to check if the database exists before running mongodump.

Unfortunately mongoDB has a function to check if a collection exists (db.collection.exists), but not to check if a databases exists.

Sure, you should only try to backup existing databases. But what will happen if something or somebody removes a database by mistake? Backup job will get stuck!

So this small piece of code will do the trick:

var found=new Boolean(false);var dbs = db.getMongo().getDBNames(); for(var i in dbs) { if (dbs[i].toString == ‘<database_name>’)  found = 1;
}
if (found.valueOf()) print(true); else print(false); 

  Download code   You only need to change <database_name> to the database name you want to check.

Of course you can use this as a function. Or you can use it inside a Bash Script used to a single database:

# Function to create a gzipped backup for a database
# Syntax: dbbackup node db_name directory
function dbbackup { local js=” var found = 0; var dbs = db.getMongo().getDBNames();
for(var i in dbs) { if (dbs[i].toString() == ‘”${2}”‘) found = 1; }
if (found.valueOf()) print(true); else print(false); “ if [ $(mongo –host $1 –eval “$js” –quiet) == “true” ]; then # Flush and lock node
mongo –host $1 –eval ‘db.fsyncLock()’ –quiet
# Dump, compress and remove dump folder
mongodump –host $1 –db $2 –out $3
tar -C $3 -czf $3/$2.tar.gz $2
rm -rf $3/$2
# Unlock node
mongo –host $1 –eval ‘db.fsyncUnlock()’ –quiet
echo “Database backed up correctly!” else echo “Database $2 doesn’t exist!” fi }

  Download code

And yes, I know making block level snapshots will be better. But sometimes you simply can’t use them.