Hi!
We have a QNX4 dual network with 50 nodes in a plant.
We discovered an interesting behavior of the network,
when one of the two networks - say logical 1 - is down.
The network manager tries to transfer through the bad
network in every 40 seconds (this is the default value
of Net -t option). It couses a 3 second delay in
data transfer (the default value of Net.* -t and -n
options is 20 - 1 second - and 3).
You can say, that this is a documented behavior of the
network manager. The big surprise is, that the network
manager tests the bad network with the real message instead
of a test message. If I were the software designer of Net,
I should use a test message for testing the link whitout
disturbing the high level communication.
Of course I have the possibility to decrease the -t and -n
options of Net.* driver, but in some circumtancies it can
couse other problems.
So my quiestions are: Is it true, that the feature of a dual
network is worse than a single network, when one of the
networks is down? Has anybody faced this problem before?
Thanks,
Jozsef
| | | | | |_ |_ || || |
|| || |_ | | | | | | |_
Jozsef Pal, Ph.D.El.Eng. Voice: +36 1 457-82-26
VEIKI Research Institute Fax: +36 1 457-82-28
Division for Control Engineering Mobil: +36 30 960-69-23
H-1251 Budapest, P.O.B. 80. Email: j.pal@veiki.hu
HUNGARY Web: http://www.veiki.hu
Jozsef Pal wrote:
On Wed, 6 Nov 2002 06:49:36 -0500, “Brown, Richard”
brownr_aecl_ca@127.0.0.1> wrote:
Yup, we are facing the same situation. We are investigating the use of
automap to work around this problem. Automap can be configured to mask down
links. Once masked Net no longer tries it.
Automap? Where can I get it from? How does it work?
Once masked out, who will it set back?
www.zinformatic.com
I wrote my own utility to do the masking when one of the networks is
down (automap is a great product, but is overkill for our requirements)
using netraw. I send “test messages” using a custom protocol, and map
out a LAN when it is down (we also set all the timeouts way down, so the
delay is not a problem, I simply need diagnostics, so that an operator
knows to fix a downed LAN).
Yup, we are facing the same situation. We are investigating the use of
automap to work around this problem. Automap can be configured to mask down
links. Once masked Net no longer tries it.
“Jozsef Pal” <j.pal@veiki.hu> wrote in message
news:3dc8c03f.1057312@inn.qnx.com…
Hi!
We have a QNX4 dual network with 50 nodes in a plant.
We discovered an interesting behavior of the network,
when one of the two networks - say logical 1 - is down.
The network manager tries to transfer through the bad
network in every 40 seconds (this is the default value
of Net -t option). It couses a 3 second delay in
data transfer (the default value of Net.* -t and -n
options is 20 - 1 second - and 3).
You can say, that this is a documented behavior of the
network manager. The big surprise is, that the network
manager tests the bad network with the real message instead
of a test message. If I were the software designer of Net,
I should use a test message for testing the link whitout
disturbing the high level communication.
Of course I have the possibility to decrease the -t and -n
options of Net.* driver, but in some circumtancies it can
couse other problems.
So my quiestions are: Is it true, that the feature of a dual
network is worse than a single network, when one of the
networks is down? Has anybody faced this problem before?
Thanks,
Jozsef
| | | | | |_ |_ || || |
|| || |_ | | | | | | |_
Jozsef Pal, Ph.D.El.Eng. Voice: +36 1 457-82-26
VEIKI Research Institute Fax: +36 1 457-82-28
Division for Control Engineering Mobil: +36 30 960-69-23
H-1251 Budapest, P.O.B. 80. Email: > j.pal@veiki.hu
HUNGARY Web: > http://www.veiki.hu
On Wed, 6 Nov 2002 06:49:36 -0500, “Brown, Richard”
brownr_aecl_ca@127.0.0.1 wrote:
Yup, we are facing the same situation. We are investigating the use of
automap to work around this problem. Automap can be configured to mask down
links. Once masked Net no longer tries it.
Automap? Where can I get it from? How does it work?
Once masked out, who will it set back?
| | | | | |_ |_ || || |
|| || |_ | | | | | | |_
Jozsef Pal, Ph.D.El.Eng. Voice: +36 1 457-82-26
VEIKI Research Institute Fax: +36 1 457-82-28
Division for Control Engineering Mobil: +36 30 960-69-23
H-1251 Budapest, P.O.B. 80. Email: j.pal@veiki.hu
HUNGARY Web: http://www.veiki.hu
Mario Charest wrote:
Out of curiousity how much time did it took you to write this utility?
To be honest about three days (to get it bug free), and then another
2 days of QA time (find the bugs). Which is a lot more than it would
have cost to buy automap. The “overkill” I speak of is that automap
does many more things than this utility (which is only capable of
automapping, and masking 2 nodes - like I said we have very special
requirements).
ps: the protocol ID is 0xDEADABBA, and it is always a completely
isolated LAN with only 2 nodes (the effort to do this vs. what
automap does is not comparable).
“Rennie Allen” <rallen@csical.com> wrote in message
news:3DC8C9A2.5030303@csical.com…
Jozsef Pal wrote:
On Wed, 6 Nov 2002 06:49:36 -0500, “Brown, Richard”
brownr_aecl_ca@127.0.0.1> wrote:
Yup, we are facing the same situation. We are investigating the use of
automap to work around this problem. Automap can be configured to mask
down
links. Once masked Net no longer tries it.
Automap? Where can I get it from? How does it work?
Once masked out, who will it set back?
www.zinformatic.com
I wrote my own utility to do the masking when one of the networks is
down (automap is a great product, but is overkill for our requirements)
using netraw. I send “test messages” using a custom protocol, and map
out a LAN when it is down (we also set all the timeouts way down, so the
delay is not a problem, I simply need diagnostics, so that an operator
knows to fix a downed LAN).
Out of curiousity how much time did it took you to write this utility?
“Rennie Allen” <rallen@csical.com> wrote in message
news:3DC932CD.7000601@csical.com…
Mario Charest wrote:
Out of curiousity how much time did it took you to write this utility?
To be honest about three days (to get it bug free), and then another
2 days of QA time (find the bugs). Which is a lot more than it would
have cost to buy automap. The “overkill” I speak of is that automap
does many more things than this utility (which is only capable of
automapping, and masking 2 nodes - like I said we have very special
requirements).
ps: the protocol ID is 0xDEADABBA,
Aren’t protocol ID 16 bit?
isolated LAN with only 2 nodes (the effort to do this vs. what
automap does is not comparable).
Mario Charest wrote:
Aren’t protocol ID 16 bit?
Yes, they are. The second 16bit is 0xABBA for fun (but it does serve a
purpose). When broadcasting on the first logical LAN the id is 0xABBA, on
the second LAN it is 0xABBB, etc.). The utility is specifically designed
to automatically maintain mappings between 2 nodes for multiple redundant
LANs and to map out defective LANs and generate alarms on our NT
workstation
software when a LAN is down.
ps: no I don’t hate ABBA, but there are only so many things you can
spell
in hex 
pps: I don’t like ABBA either…
“Rennie Allen” <rallen@csical.com> wrote in message
news:aqe30j$ika$1@tiger.openqnx.com…
ps: no I don’t hate ABBA, but there are only so many things you can
spell
in hex >
pps: I don’t like ABBA either…
I always liked the QNX2 special addr 0xCAFEFACE.