Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[brandx]

I would love to see the option to be able to have 2 different services box's running. one as a primary, and one for backup that just replicates the databases with the primary unless the event arises that the primary goes down..

I have seen many networks taken down just by taking out the hub box.  So admins will alot of times configure a backup hub.. well this would be a similar feature...

[Vorex]

Originally posted by brandx
I would love to see the option to be able to have 2 different services box's running. one as a primary, and one for backup that just replicates the databases with the primary unless the event arises that the primary goes down..

I have seen many networks taken down just by taking out the hub box.  So admins will alot of times configure a backup hub.. well this would be a similar feature...

SCP the db's to another box with services preconfigured when the need arises.

[Kei]

It sounds like a good idea. Im not sure how we could implement it.... Maybe a config directive isBackup or something.... as for running, I think we can define something globally to not introduce clients until the Primary goes offline.

[Pieter Bootsma]

The most important question is how you're going to sync databases, and it has to be in such a way that netsplits on the net or in a direct link between services don't break the consistency of the databases.

[Kei]

No solution would ever be perfect in this case - as a SCP/SSH manual backup would not even be perfect for running a second copy of services.... if your services' box goes down, poof. None of the nets I know make backups on a regular basis lol

That said, the Master copy of services could transmit information to the Backup copy of services via IRC. (One would hope, that you'd supply your Backups with your DB files before running them the first time, to reduce bandwidth usage on linkup).

One hitch to all of this though, that I just thought of... especially with Unreal servers. Since your two services daemons cannot have the same server name (Services.yournet.net) The backup will have to be differently named. This will definitely cause issues when using /ns /cs /ms /os -- and break Services-dependent commands, as the IRCd will think services are offline.

[ShelLuser]

We're still in the process of setting such a thing up ourselves and to this end we (ab)used a VPN project which we had setup several years ago. So the main hubs of our network are also linked through a VPN which allows them to keep the MySQL data synchronized. This is a one way connection ofcourse; the "backup hub" basicly runs an "sqldump" on the database and imports this into its own.

Next we've been planning to run the extra services in read/only mode with both the default and alternative nicks set. While we haven't tested this yet it is our understanding that we'll end up with a 2nd set of services which use other nicks. When the first services go down these nicks will change and your services are being maintained.

To my knowledge (not sure here!) this is already supported by Anope. The only possible issue you might have here is when the services come back up again. Its basicly where our investigation in this has ended ;-)