odd pg_start_backup() issue

Hi All;

I'm setting up PITR based backups on a production db server.
When the script runs the pg_start_backup it takes forever, like 30minutes+

I can run a checkpoint while waiting for it to complete and it finishes
almost immediately

For now we're executing the start backup like this (in the automated
base backup script:
echo "checkpoint; select pg_start_backup('$tag'); checkpoint;" | psql

we tried it like this and id did not help:
echo "checkpoint; select pg_start_backup('$tag');" | psql
so we added the checkpoint after the pg_start_backup command

Next I'll try this (I assume the first checkpoint is simply un-needed):
echo "select pg_start_backup('$tag'); checkpoint;" | psql

However, this works well and the pg_start_backup finishes in less than
1minute,
but if I leave the checkpoint command(s) out then we're back to
30minutes or more for the pg_start_backup...

We're running version 9.2.4 on a Centos 6 server

Thoughts?

Thanks in advance