I'd like to suggest a change in the way that SQL servers and backups are
treated in radiator.  What I'd like to see is the FailureBackoffTime
represent the amount of time that an SQL servers is to be not be contacted
again and the backup used.  As it stands now, if the primary server
doesn't respond, the backup SQL server is used until it times out and then
it moves back through the list of db's to contact.  The behaviour I'd like
to see is that the backup server is used when the primary doesn't respond
until FailureBackoffTime is reached - then the primary is recontacted.  
If it responds then the process starts over again.  Right now the
secondary/backup would take all requests forever or until it times out and
then the list is retried.  Since many db's have a cleanup routine where it
can become unavailable for a short amount of time this behaviour would
make more sense to me.  You could tune FailureBackoffTime to be around the
length of time your cleanup job takes so that the backup server would get
your through that period.  The main issue I've got is that you don't
really know that the requests are going to the secondary/backup server and
it may stay there for quite some time degrading performance (assuming that
your primary db is setup to give better response due to location, machine
type, etc.)  Thoughts?

--------------------------------------------------------------------------
Aaron Holtz
ComNet Inc.
UNIX Systems Administration/Network Operations
"It's not broken, it just lacks duct tape."
--------------------------------------------------------------------------



===
Archive at http://www.thesite.com.au/~radiator/
To unsubscribe, email '[EMAIL PROTECTED]' with
'unsubscribe radiator' in the body of the message.

Reply via email to