new-slave.py should refuse to run if cluster is lagged

[Stuart Bishop]

If the cluster is lagged, there may be a problem. Even if there is not a problem, the script will block until lag is cleared. We should just abort if the cluster is lagged so the lag can be diagnosed and corrected.

[Stuart Bishop (stub)]

Also check for events to or from the new node_id in sl_event and sl_confirm (these might be laying around if the cleanup thread hasn't been run since a node with the same id has been dropped).

[Stuart Bishop (stub)]

Checking for long running transactions on all nodes good too.

Ideally all tools in database/replication get these checks.

[Gary Poster (gary)]

We would like this to be done within the week, so it can be part of the last 2009 release in two weeks. Moving to "High".

[Gary Poster (gary)]

Stuart reports that this should no longer be a problem once we are able to no longer do the auth database replication. This will greatly simplify our scripts.