This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.
>From: Erick Lorenz <address@hidden> >Organization: UC Davis >Keywords: 200009220039.e8M0dBb07890 McIDAS-X 7.60 Linux Erick, re: multiple announcements >I remember those announcements but they didn't sink in that they applied to >the AREA files. I think I was hoping that since I was getting the latest >version that all this would be in place. >>redirect.k ADD ROUTE.SYS \"/var/data/mcidas >>batch.k CIMSS.BAT > >I had to do a ROUTE INIT first before this worked. I had assumed tht since you were attempting to use lwtoa3 that you had already put a copy of a viable ROUTE.SYS into the target directory, /var/data/mcidas in your case. The addition of the CIMSS products to the routing table would then be an extra. The fact that you did not have to do a ROUTE INIT tells me one of two things: o your REDIRECTion to the copy of ROUTE.SYS in the output data directory did not exist o you had a valid REDIRECTion, but no routing table >Anyway as soon as I did all this, AREA0000 popped up in /var/data/mcidas >before I could do an ls on that directory so I think everything is fine. Unfortunately, everything is not fine. A file named AREA0000 indicates a problem. The reason? AREA0000 is an invalid AREA file name. What you should do is: <as 'ldm'> ldmadmin stop <as 'mcidas' make sure that the LDM is not running when you do this> cd workdata cp ROUTE.SYS /var/data/mcidas redirect.k ADD ROUTE.SYS \"/var/data/mcidas batch.k CIMSS.BAT <as 'ldm'> ldmadmin start >Perhaps I should wait to celebrate but I think this a record for the fewest >number of "HELP!"s on an installation for me. You are pretty close, but not out of the woods yet. >I am now going to set up some user accounts that have home directories on >ATM20, the server, even though the users will be logging in from other >Linux stations. I want their home directories to be on the server because >it is the only host that will be running Linux all the time and besides the >users will not always be able to get to the same station. > >What I am now wondering is whether it is best to copy the mcidas manager >directory to each client host so that the user is running mcidas off of >local binaries or whether it would work to export the mcidas manager from >ATM20 to all the clients. The simplest setup is for each machine to NFS mount the file system where McIDAS-X is installed. Then each user's account should include ~mcidas/bin (e.g. /home/mcidas/bin or where ever the McIDAS binaries are located) in his/her PATH. Also, each user will need to have McIDAS environment variables defined for his/her session. This is done in .cshrc (C shell users; .profile for Bourne/Korn shell users). What has to be defined is described in detail in: Configuring Unidata McIDAS-X Accounts http://www.unidata.ucar.edu/packages/mcidas/mcx/config_mcx.html >1. ATM20 (server) > /home/mcidas (for use on 20 only) > /home/users/mcidas-user1 > /home/users/mcidas-user2 > ATM51 (client) > /usr/local/mcidas (for use by ATM51) > /home -> ATM20:/home (nfs mount) > /mcidas-user1 -> atm20:/home/users/mcidas-user1 > /mcidas-user2 -> atm20:/home/users/mcidas-user2 > ATM52 (client) > /usr/local/mcidas (for use by ATM52) > /home -> ATM20:/home (nfs mount) > /mcidas-user1 -> atm20:/home/users/mcidas-user1 > /mcidas-user2 -> atm20:/home/users/mcidas-user2 > >2. ATM20 (server) > /home/mcidas (for use by all machines) > /mcidas-user1 -> atm20:/home/users/mcidas-user1 > /mcidas-user2 -> atm20:/home/users/mcidas-user2 > ATM51 (client) > /home -> ATM20:/home (nfs mount) (sees /home/mcidas) > /mcidas-user1 -> atm20:/home/users/mcidas-user1 > /mcidas-user2 -> atm20:/home/users/mcidas-user2 > ATM52 (client) > /home -> ATM20:/home (nfs mount) (sees /home/mcidas) > /mcidas-user1 -> atm20:/home/users/mcidas-user1 > /mcidas-user2 -> atm20:/home/users/mcidas-user2 > >My guess is that option 2 might have poorer performance even though users >are running mcidas with their own CPUs just because the load times are >dependent on network traffic but it would also be much easier to maintain. > >If (2) is possible is the trade-off worth it? Performance of NFS on Linux is known to be _not so good_ (but improving with time), so it might be better to replicate the existence of the ~mcidas directory strucuture on each machine. I want to add, however, that if there will be simultaneous McIDAS-X sessions run from different machines by the same user, then you will run into problems with the sessions stepping on each other. From your listing I think I am seeing that you want to have 4 accounts setup for McIDAS-X use. Is this correct? Are you planning on having more that 4 McIDAS-X sessions active at the same time? Is your planned setup based on having a few user accounts that are accessible by a large number of sessions (again, having multiple McIDAS-X sessions running out of the same account from multiple machines)? McIDAS works best (has the least number of potential gotchas) if each McIDAS-X user runs his/her McIDAS-X session from their own account. The overhead for doing this is small. Each user simply needs to have his/her own mcidas/data subdirectory. The McIDAS-X binaries are shared from ~mcidas/bin; the ancillary data files are shared from ~mcidas/data (ancillary data files are things like map outline files, enhancements, stretch tables, etc.); and data from /var/data/mcidas (your system setup from above). As soon as you try to run simultaneous McIDAS-X sessions from a single account but on different machines you _will_ run into problems unless you know exactly what you are doing (and students never really know what they are doing). If you give me a better idea of exactly what you are trying to accomplish, I can help you layout a setup that should work nicely. Tom