Information on Stopping/Starting Distributed Systems

There are no special rules to take into account when stopping a distributed system. You can stop TREX in any order on the individual hosts.

When you start a distributed system, the type of data storage dictates whether there is a defined sequence.

• If you are using centralized data storage, there is no special sequence.

• If you are using decentralized data storage, you firstly have to start a master name server that was running just before the system was stopped. This ensures that the system is based on an up-to-date topology file.

  Example

  The hosts mytrexhost1, mytrexhost2, and mytrexhost3 are configured as master name servers. mytrexhost3 has not been operating for a while, which means that its topology file is not up-to-date. Changes that have been made since (such as new indexes) are not known to this host.

You also stop TREX on the remaining hosts for maintenance reasons. If you now want to restart TREX, you now have to start it on mytrexhost1 or mytrexhost2 first. These master name servers have up-to-date topology files.

If you were to start TREX on mytrexhost3 first, the system would be based on an out-of-date topology file.