This file explains how to operate a Zephyr service once you have installed Zephyr on all the relevant machines and file servers in your environment. To learn how to configure, build, and install Zephyr, read the file INSTALL. To set up Zephyr service at a site, follow these steps: 1. Choose the machines you wish to have act as Zephyr servers at your site. Expect the server to be CPU-efficient but to consume a bit more memory than you might expect (at MIT, with around a thousand simultaneous users, the Zephyr server's data size is 40MB). If you have a lot of users, the server should have enough memory so that the process doesn't swap. 2a. If you configured Zephyr with Hesiod support, make sure your Hesiod realm has a "zephyr.sloc" entry containing a record for each server. (Each entry should contain the name of the server, nothing else.) The Zephyr servers will use the zephyr.sloc entry to find the other servers. Host managers will use the zephyr.sloc entry to find the Zephyr servers by default; however, you can control the set of servers for each host manager by giving each host a ".cluster" entry containing a record "zcluster ". If such a record is found, the host manager will resolve ".sloc" instead of "zephyr.sloc". 2b. If you configured Zephyr without Hesiod support, and you have multiple Zephyr servers, each server should have a file "server.list" in the configuration directory (which is /etc/athena/zephyr if you configured with --enable-athena, or /usr/local/etc/zephyr if you installed Zephyr in /usr/local and didn't use --enable-athena). This file should contain a list of the servers, one per line. 3. If you configured Zephyr with Kerberos 4 support, make a service key "zephyr.zephyr@" and install a srvtab for that service as "srvtab" in the configuration directory of each of your zephyr servers. 4. Start zephyrd from the system binary directory (/usr/athena/etc if you configured with --enable-athena, /usr/local/sbin if you installed in /usr/local and didn't use --enable-athena). zephyrd logs as service "local6"; watch the syslogs for error messages. Arrange for zephyrd to be run at boot time on your server machines. 5. Each client machine should run zhm (the Zephyr Host Manager) from the local system binary directory (/etc/athena for --enable-athena, /usr/local/sbin if you installed in /usr/local and didn't use --enable-athena). If you built Zephyr without Hesiod support, you should start zhm as "zhm server1 server2 server3 ..." so that zhm knows where the Zephyr servers are. Do not use "localhost" or "127.0.0.1" as a server name, or zhm will become confused. You can send a SIGFPE signal to the server process to make it dump its subscription database to /var/tmp/zephyr.db. (If /var/tmp didn't exist when Zephyr was built, the subscription database will be dumped in /usr/tmp or /tmp instead.)