by Doug Robinson
Very good presentation, with convincing solutions to problems we meet. It handles the replication of multiple vobs all over the world with no down-time, minimum load on vob servers and optimized network congestion.
Doug has worked tightly with the ClearCase developers for long (was previously working for Atria, one a of the two original designers of DSEE -- the other was David Leblang), and is very respected. ClearCase people let him answer some questions for them during BOFs.
One large project.
One central backup site, all replicas replicated there, and the
disaster recovery handled there.
It is the one site that has preserving replicas (for
backup/recovery purposes). One NIS per site.
Only one user group per site.
Enabling setuid is an issue. Avoided as much as possible.
Doug had a presentation last year on recursive labelling. He recommends the use of time based labels, for scalability reasons.
Rational provided scripts are not optimized. They do not make use of all the features supported by the shipping server.
There is one central shipping server forwarding hub.
In case of DNS failure, the packets are returned. Therefore, there is a DNS cache on all vob and shipping servers.
Scrub the oplogs eventually. Use the -omaster option, e.g. to mkattr.