rabbe@infolog.se (Rabbe Fogelholm) (02/15/91)
We are considering an extension of our Domain file system, but I wonder
if it is really feasible.
Today we have two disked and a few diskless nodes on a Token Ring. One
of the disked nodes is also connected to a general-purpose local
Ethernet so that we can do ftp to Suns and the like.
First of all, would it be possible to set up another disked DN3000,
connect it to the Ethernet, and thus have it become a member of the
existing Domain file system? I. e., users on the Token Ring nodes
should be able to access the new disk by descending from the network
root (//) as usual.
Second, we would like to move the new DN3000 to another city and put
some suitable communications link between the DN3000 and our local
Ethernet. This should be invisible to the users, except that
slower-than-usual file accesses would be tolerable.
Any comments are welcome. If the idea seems sound, what kind of
communications link should we use?
(Some technical details: We run 9.7 today but plan for 10.3 soon. The
node that speaks Token Ring and Ethernet at the same time is a DN4500.
The rtsvc command currently returns the following:
Controller Net ID Service offered
================== ======== ====================
RING 0 Own traffic only
ETH802.3_AT 0 Own traffic only )krowitz@RICHTER.MIT.EDU (David Krowitz) (02/15/91)
Apollo nodes that are on differnet physical network *may* share the same Domain file system *IF* the network gateway is an Apollo workstation with "rtsvc" enabled to perform routing *OR* if the network gateway is a machine which accepts the Apollo DDS network protocal (I thinkt the Cisco routers do this?). The Apollo file system does *not* use TCP/IP network protocals (they are too slow) and can not span an internet in which one or more or (of) the routers is a TCP/IP-only router. Apollo workstation can run NFS in addition to the Apollo file system, and two or more Apollo workstations *can* share files across an internet via NFS, although NFS has certain limitations (ie. Apollo file-types are not visible across NFS links). I believe that Apollo had a product which would connect two remote Apollo sites via a T1 link which allowed the full Apollo DDS networking protocals to be used, but a leased T1 line is kind of expensive for a single node. -- David Krowitz krowitz@richter.mit.edu (18.83.0.109) krowitz%richter.mit.edu@eddie.mit.edu krowitz%richter.mit.edu@mitvma.bitnet (in order of decreasing preference)
thompson@PAN.SSEC.HONEYWELL.COM (John Thompson) (02/16/91)
> <<forwarded message>> > > Today we have two disked and a few diskless nodes on a Token Ring. One > of the disked nodes is also connected to a general-purpose local > Ethernet so that we can do ftp to Suns and the like. > > First of all, would it be possible to set up another disked DN3000, > connect it to the Ethernet, and thus have it become a member of the > existing Domain file system? I. e., users on the Token Ring nodes > should be able to access the new disk by descending from the network > root (//) as usual. Piece of cake (yes, even at 9.7). Some commands, such as 'lcnode', will not list all the nodes, because those commands are limited to a single net. However, all the transparent network stuff you know and love can bridge. I'd suggest reading the (10.x) manual _Managing_Domain_O/S_And_Domain_Routing_ In_An_Internet_ (005694-A00) as a guide. There's an equivalent 9.7 manual too (though not much has changed). Basically, it goes as follows: 1) Choose network IDs for each physical net. These are hex #s, and can be 5 digits long, I know (beyond 5, I don't guarantee). 2) On each node (lucky you have only a few), execute the command '/etc/rtsvc -dev ring -net 12345' to change it to net 12345. This will set the netID for all time (and will also confuse the rest of the ring until the other nodes get re-set too. If you're at sr10, I wouldn't be surprised if you need to reboot at least the rgyd and glbd nodes. 3) On the bridging-node, add '-route' to the tail of the previous command, and also enter '/etc/rtsvc -dev eth802.3_at -net 54321 -route' to let it know the netID of the ethernet, and to tell it to offer routing between nets 4) On the ethernet node(s), use the rtsvc command to set those netIDs. 5) On each node of the ring, do a 'ctnode ethernode 54321.NODEID' to tell each node about the ethernet node(s) that are there. Do the opposite (ctnode ringnode 12345.NODEID) on the ethernet node(s). 6) Instead/in-addition-to (5), run a name-server helper (ns_helper) to offer name-service resolution. It should probably run on the gateway, so that both sides can see it directly. Read the manual for how to initialize it etc. NOTE: When we went to sr10 (and re-involed the disks), the netID was lost. This wouldn't be so bad except that, when booted up in the net, it could see the rest of the nodes just well enough to tell that there was a registry out there, but not well enough to get at it. As a result, logins hung FOREVER. We got around it by doing an initial boot out of the net, logging in as user.server, and running rtsvc to set the netID (and properly name 'node_XYZZY'). I have been assured that the hint_file will eventually clear everything up, but I waited over 2 hours for it to get unconfused, without any luck. Now I spend about a minute clearing it by hand. > Second, we would like to move the new DN3000 to another city and put > some suitable communications link between the DN3000 and our local > Ethernet. This should be invisible to the users, except that > slower-than-usual file accesses would be tolerable. You need to be careful about what bridges/routers are in between. The link needs to be pretty transparent. Apollo claims you needed a full T1 speed link in order for internet DDS to work. We ran it out to Colorado (from Minnesota) on (I believe) a 56Kb line. I've been told that it can 'tolerate' speeds down to 19.2Kb, as long as the load is light. (Note that HP/Apollo won't support a slower line if you have problems on it....) -- jt -- John Thompson Honeywell, SSEC Plymouth, MN 55441 thompson@pan.ssec.honeywell.com Me? Represent Honeywell? You've GOT to be kidding!!!
nazgul@alphalpha.com (Kee Hinckley) (02/17/91)
In article <9102151426.AA22291@richter.mit.edu> krowitz@RICHTER.MIT.EDU (David Krowitz) writes: >Apollo nodes that are on differnet physical network *may* share the same >Domain file system *IF* the network gateway is an Apollo workstation with >"rtsvc" enabled to perform routing *OR* if the network gateway is a machine >which accepts the Apollo DDS network protocal (I thinkt the Cisco routers >do this?). The Apollo file system does *not* use TCP/IP network protocals >(they are too slow) and can not span an internet in which one or more or >(of) the routers is a TCP/IP-only router. Apollo workstation can run NFS I assume this means you couldn't use a Telebit NetBlazer? -- Alfalfa Software, Inc. | Poste: The EMail for Unix nazgul@alfalfa.com | Send Anything... Anywhere 617/646-7703 (voice/fax) | info@alfalfa.com I'm not sure which upsets me more; that people are so unwilling to accept responsibility for their own actions, or that they are so eager to regulate everyone else's.