This happens because when there are multiple routing nodes for the same
area, DECnet uses the one with the highest address. The obvious solution
would be for Johnny to renumber MIM as, say, 1.1023.
Hmm.
CTAKAH is set to lower router priority than MIM. I thought this is enough to tell that MIM is preffered router...
There's no need for CTAKAH to be and area router, so I'll downgrade it to endnode...
I wrote-
OpenVMS Network status for local node 2.1 LEGATO on 5-JAN-2012
08:11:19.43
Area Cost Hops Next Hop to Area
1 4 1 QNA-1 -> 1.300 CTAKAH
....
It bugs me, although I can't point at an actual problem that it causes,
that the routing node for area 1 is seen as CTAKAH, rather than MIM. Since
CTAKAH is not actually local to Uppsula (I think that's where MIM is) but is
rather at the other end of yet another bridge, this means that any traffic
for area one, no matter what its ultimate destination, has to make an extra
round trip to Russia.
This happens because when there are multiple routing nodes for the same
area, DECnet uses the one with the highest address. The obvious solution
would be for Johnny to renumber MIM as, say, 1.1023.
Bob
On 5 Jan 2012, at 15:58, "Steve Davidson" <jeep at scshome.net> wrote:
It could be the issue. If you are running Phase-IV then NETCONFIG will allow you to create the necessary account(s). You have a choice of creating a default DECnet account, or creating a FAL$SERVER account, or both. I have both contrary to what DEC's Inspect software would prefer.
Thanks for the heads-up I'll sort that out this evening.
What is the INFO.TXT you guys are talking about? Should I have one on STAR69 with my node list in?
HECnet is a closed society, so I don't worry about it.
Ha, wasn't a security issue as much as a clueless operator issue ;)
--
Mark Benson
http://markbenson.org/bloghttp://twitter.com/MDBenson
where DECnet *might* get confused.
Well, this is easy enough to test experimentally. I've disabled the
redundant Multinet links on LEGATO, so in theory there should be no
redundancy now at least as far as LEGATO goes.
This means not only disabling the direct links from LEGATO to other bridge
nodes (e.g. GORVAX, SG1) but _also_ the links from LEGATO to other nodes
that in turn have Multinet links to bridge nodes (STUPI, FRUGAL, ROOSTA).
In theory I suppose the latter step is excessive since we can assume that a
two hop path would always cost more than a one hop path, but there's no
guarantee of that. So just to be safe this absolutely eliminates redundant
paths.
I think that leaves me with area 54, CIERE, as the only machine that
_only_ gets its connectivity thru LEGATO. That link I've left turned on.
Now LEGATO sees the path to everything except CEIRE as being thru the
bridge (QNA-1), which is what we'd expect.
OpenVMS Network status for local node 2.1 LEGATO on 5-JAN-2012
08:11:19.43
Area Cost Hops Next Hop to Area
1 4 1 QNA-1 -> 1.300 CTAKAH
2 0 0 (Local) -> 2.1 LEGATO
3 7 2 QNA-1 -> 19.41 SG1
4 4 1 QNA-1 -> 4.249 SLAVE
5 14 4 QNA-1 -> 19.41 SG1
6 4 1 QNA-1 -> 6.1 STAR69
8 7 2 QNA-1 -> 19.41 SG1
11 4 1 QNA-1 -> 11.2 MAISA
19 4 1 QNA-1 -> 19.41 SG1
20 14 2 QNA-1 -> 19.41 SG1
33 7 2 QNA-1 -> 19.41 SG1
42 4 1 QNA-1 -> 42.1 CANADA
52 10 3 QNA-1 -> 19.41 SG1
54 2 1 TCP-0-54 -> 54.59 CEIRE
59 7 2 QNA-1 -> 19.41 SG1
Try your file copies again and see if anything changes...
Bob
It could be the issue. If you are running Phase-IV then NETCONFIG will allow you to create the necessary account(s). You have a choice of creating a default DECnet account, or creating a FAL$SERVER account, or both. I have both contrary to what DEC's Inspect software would prefer. HECnet is a closed society, so I don't worry about it.
-Steve
________________________________
From: owner-hecnet at Update.UU.SE on behalf of Mark Benson
Sent: Thu 1/5/2012 10:55
To: hecnet at Update.UU.SE
Subject: Re: [HECnet] Multinet circuit costs
On 5 Jan 2012, at 15:40, hvlems at zonnet.nl wrote:
Yes I have problems wihile copying files from Legato and Star69 to Ozon.
STAR69 has no public user account, and IIRC won't allow file access via DECnet without a username and password. I haven't got around to figuring out how to set up non-privileged file access in a safe directory yet.
Could this be the issue? I really am a total newbie so STAR69's capabilities are the absolute minimum needed to do it's job at present.
--
Mark Benson
http://markbenson.org/bloghttp://twitter.com/MDBenson
The circuit costs here are set to 4 for LAN and 3 for tunnels. The issue may be that if a node (at the end of the chain so to speak) has 2 tunnel links of the same cost and that the final destination costs the same I could see where DECnet *might* get confused. Let's look at the following contrived example:
FRUGAL:: connects to LEGATO:: - cost of 3
FRUGAL:: connects to SG1:: - cost of 3
LEGATO:: connects to GORVAX:: - cost of 3
SG1:: connects to GORVAX:: - cost of 3
Fred, on node FRUGAL:: tries to copy a (large) file from GORVAX:: to FRUGAL:: *may* see some issues. On the other hand if Fred chooses to favor SG1:: over LEGATO:: (cost wise) he should not see any problems: Ex:
FRUGAL:: connects to LEGATO:: - cost of 3
FRUGAL:: connects to SG1:: - cost of 2
The route is guaranteed to be the same for the entire transaction. It will be (using poor man's routing to demonstrate) FRUGAL::SG1::GORVAX:: at a cost of 5, not FRUGAL::LEGATO::GORVAX:: at a cost of 6.
-Steve
________________________________
From: owner-hecnet at Update.UU.SE on behalf of hvlems at zonnet.nl
Sent: Thu 1/5/2012 10:43
To: hecnet at Update.UU.SE
Subject: Re: [HECnet] Multinet circuit costs
PS
I can copy to Legato alright.
PPS
I did check the netserver.log files and found nothing out of the order.
On 5 Jan 2012, at 15:40, hvlems at zonnet.nl wrote:
Yes I have problems wihile copying files from Legato and Star69 to Ozon.
STAR69 has no public user account, and IIRC won't allow file access via DECnet without a username and password. I haven't got around to figuring out how to set up non-privileged file access in a safe directory yet.
Could this be the issue? I really am a total newbie so STAR69's capabilities are the absolute minimum needed to do it's job at present.
--
Mark Benson
http://markbenson.org/bloghttp://twitter.com/MDBenson
Again, apologies for topposting: my cellphone is somewhat rude...
PST is UTC-8 (unless Bob lives in the Philippines, then it's UTC+8 :).
Yes I have problems wihile copying files from Legato and Star69 to Ozon.
Ozon is the area router for area 44 and uses the bridge software.
I can't access the Hecnet map but Legato has multiple paths and IIRC so does Star.
Ozon has no problem copying files from MIM or MUSIPC and both are connected with single paths (IIRC again).
Reading Bob's analysis I still think that multiple paths may lead to errors. Granted that circuit cost may fix this but right now it is, for me at least, not clear.
The area information in the output of show network/old on Ozon shows all areas and about 40% is more than 1 hop away and seen as reachable thru Legato.
Hans
-----Original Message-----
From: "Bob Armstrong" <bob at jfcl.com>
Sender: owner-hecnet at Update.UU.SE
Date: Thu, 5 Jan 2012 07:14:50
To: <hecnet at Update.UU.SE>
Reply-To: hecnet at Update.UU.SESubject: RE: [HECnet] Multinet circuit costs
Why? What is the percieved problem?
I don't know about area 44, but I was talking to Hans on New Years Eve
about a problem where he could not copy files, while logged in to OZON, from
LEGATO (i.e. a "pull" transaction). According to Hans, it fails with "acp
file or directory lookup failed". If I did the same thing from my end (a
"push") it would start copying and actually create an empty file on OZON,
but then time out after a very long time with
-RMS-F-SYS, QIO system service request failed
-SYSTEM-F-LINKABORT, network partner aborted logical link
Last I heard Hans was going to check the NETSERVER/FAL logs on his side for
any error messages, but that was New Years Eve and he was otherwise occupied
:-)
I see this morning that Hans has actually created a file "TARGET.TXT" on
LEGATO (dated 6PM PST on 12/31 even - what time is that in the
Netherlands??) so maybe he's solved the problem. If he did he didn't tell
me what he did.
Hans also mentioned to me that he was having similar problems with the
node STAR69 and CANADA. I think all this came up when Hans was trying to
write his HECnet crawler script to gather INFO.TXT files...
I don't think this is related to the Multinet/Bridge configuration, but I
can't guarantee it. One hidden "gotcha" in DECnet is that the circuit costs
are determined locally, so the routings don't have to be symmetrical (i.e.
the route from A->B might not be the same as the route from B->A).
If somebody (Johnny??) wants to try to figure it out, be my guest -
GUEST/GUEST. I'm a little busy at the moment and don't really have time to
look at it.
Bob
Why? What is the percieved problem?
I don't know about area 44, but I was talking to Hans on New Years Eve
about a problem where he could not copy files, while logged in to OZON, from
LEGATO (i.e. a "pull" transaction). According to Hans, it fails with "acp
file or directory lookup failed". If I did the same thing from my end (a
"push") it would start copying and actually create an empty file on OZON,
but then time out after a very long time with
-RMS-F-SYS, QIO system service request failed
-SYSTEM-F-LINKABORT, network partner aborted logical link
Last I heard Hans was going to check the NETSERVER/FAL logs on his side for
any error messages, but that was New Years Eve and he was otherwise occupied
:-)
I see this morning that Hans has actually created a file "TARGET.TXT" on
LEGATO (dated 6PM PST on 12/31 even - what time is that in the
Netherlands??) so maybe he's solved the problem. If he did he didn't tell
me what he did.
Hans also mentioned to me that he was having similar problems with the
node STAR69 and CANADA. I think all this came up when Hans was trying to
write his HECnet crawler script to gather INFO.TXT files...
I don't think this is related to the Multinet/Bridge configuration, but I
can't guarantee it. One hidden "gotcha" in DECnet is that the circuit costs
are determined locally, so the routings don't have to be symmetrical (i.e.
the route from A->B might not be the same as the route from B->A).
If somebody (Johnny??) wants to try to figure it out, be my guest -
GUEST/GUEST. I'm a little busy at the moment and don't really have time to
look at it.
Bob
