WALLIS@TWSUVAX (Thomas Wallis) (02/08/90)
Dear Netlanders: I have in place here at Wichita State University a VAXcluster of 1 8650 (VMS 5.2) and 6 Vaxstation 3100s (diskless, VMS 5.2, boot off of the 8650). Right now, the 8650 is on BITNET and Internet. I am having reasonably good luck getting PMDF to run on the 8650. Each of the machines on the cluster will have it's own Internet address and will speak CMU TCP/IP as well as DECNET. At the moment, users of the workstations cannot get BITNET or Internet mail on their workstations, even though the mail is cluster-wide. I want to give the users the opportunity to send, receive, and reply to Internet (and BITNET) mail using their workstations. To do this, I think I must have each machine run PMDF seperately, and have different config files for each node on the cluster. I am confident I can get CMU TCP running on each VAX, but I am not sure how to set up PMDF to do the MAIL part of the equation. Has anyone out there in netland done this already? Any hints would be appreciated. Thanks in advance, Tom Wallis System Manager, Computer Science Dept. Wichita State University BITNET: WALLIS@TWSUVAX UUCP: wallis@wsucsa.UUCP or uunet!ncrlnk!ncrwic!wsucsa!wallis Internet: wallis@wsuiar.wsu.ukans.edu or wallis@wsucsa.wsu.ukans.edu
NED@HMCVAX.CLAREMONT.EDU (Ned Freed, Postmaster) (02/09/90)
Thomas Wallis writes: > I have in place here at Wichita State University a VAXcluster of > 1 8650 (VMS 5.2) and 6 Vaxstation 3100s (diskless, VMS 5.2, boot off of the > 8650). Right now, the 8650 is on BITNET and Internet. I am having reasonably > good luck getting PMDF to run on the 8650. Each of the machines on the > cluster will have it's own Internet address and will speak CMU TCP/IP as well > as DECNET. At the moment, users of the workstations cannot get BITNET or > Internet mail on their workstations, even though the mail is cluster-wide. > I want to give the users the opportunity to send, receive, and reply to > Internet (and BITNET) mail using their workstations. To do this, I think I > must have each machine run PMDF seperately, and have different config files > for each node on the cluster. I am confident I can get CMU TCP running on > each VAX, but I am not sure how to set up PMDF to do the MAIL part of the > equation. Has anyone out there in netland done this already? Any hints would > be appreciated. Question: Is this a homogeneous cluster? (Most are.) In other words, does the entire cluster share a common SYSUAF.DAT file? If so, you can get by with a single copy of PMDF and a single configuration file. You will need to install PMDF on all nodes of the cluster, and add rewrite rules for all the systems that rewrite to the local channel (preferably all of them should rewrite to the same name, e.g. the name of the boot node). The PMDF queue should run on the system with BITNET and Internet software installed (you can get clever and use multiple queues to spread the load around, but this is an advanced topic). You can set up CMU TCP/IP to invoke the PMDF slave program on all nodes. This way mail delivered to any node will be processed for all. That's really all there is to it. Mail will arrive by whatever means and enter the common PMDF queue. From there it will be delivered via the PMDF queue either to one of the outgoing networks, or via VMS MAIL. But when VMS MAIL is used it will work correctly because the user base spans the entire cluster. I run a slightly more complex variation of this theme (I have a homogeneous cluster with one hetero node (ymir) that acts as a special mail/network server). I have a single copy of PMDF for all the main cluster nodes. ymir shares PMDF executables, but has its own configuration file and queues. This setup works quite well for me. Hope this helps. Ned