QNX 4 slow after node shutdown

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I fix this?

Thanks

Augie

What aspect of performance got slower?


Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I fix this?

Thanks

Augie

“Richard R. Kramer” <rrkramer@kramer-smilko.com> wrote in message
news:3AC52FB3.A005B972@kramer-smilko.com

What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take 3-6
seconds or more to get loaded. When everything is okay, Vedit Plus brings
up the files instantaneously.

TIA

Augie

Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I fix
this?

Thanks

Augie

The same sort of thing happens here when editing across the net.
Our node 2 is usuall missing, so I don’t know how fast vedit would
be with it present. Masking node 2 makes no noticeable difference.
Commands such as ls, on the other hand, are fast.

Maybe a vedit problem?

Richard

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take 3-6
seconds or more to get loaded. When everything is okay, Vedit Plus brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I fix
this?

Thanks

Augie

We met this situation too often and our opinion is this could happen if
nameloc is running on each node of your network.
And when one node gets down, nameloc’s node polling makes the other nodes
slower.


“Richard R. Kramer” <rrkramer@kramer-smilko.com> ???/??? ? ???
???: news:3AC7CF71.147134CB@kramer-smilko.com

The same sort of thing happens here when editing across the net.
Our node 2 is usuall missing, so I don’t know how fast vedit would
be with it present. Masking node 2 makes no noticeable difference.
Commands such as ls, on the other hand, are fast.

Maybe a vedit problem?

Richard

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take 3-6
seconds or more to get loaded. When everything is okay, Vedit Plus
brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop
on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I
fix
this?

Thanks

Augie

Hi,

had a similar effect with my notebook after I took it home to work there (could
by a sign: don’t take your work home).
Running only QNX was normal, no problem.
After starting photon and launching a pterm shell, I thought something terrible
had happened to my notebook. I waited 10 or 20 seconds for appearing of the
window. That was the same with other apps from the desktop bar.
Slaying the network made it work normal.
Back in office and connected to our server all was fine.
Next time home again slow.
After some thinking I came to the symbolic links I have on my notebook. At root
level there are links to directories risiding in the server.
When I remove those links after cutting the network connection photon apps are
happy again and start as fast as with network.

Friedhelm Schuetz
H.Kleinknecht & Co. GmbH

Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I fix this?

Thanks

Augie

Hi.

“Friedhelm Schuetz” <Friedhelm.Schuetz@kleinknecht.de> wrote in message
news:3AC83751.85CF381A@kleinknecht.de

Hi,

had a similar effect with my notebook after I took it home to work there
(could
by a sign: don’t take your work home).
Running only QNX was normal, no problem.
After starting photon and launching a pterm shell, I thought something
terrible
had happened to my notebook. I waited 10 or 20 seconds for appearing of
the
window. That was the same with other apps from the desktop bar.
Slaying the network made it work normal.
Back in office and connected to our server all was fine.
Next time home again slow.
After some thinking I came to the symbolic links I have on my notebook. At
root
level there are links to directories risiding in the server.
When I remove those links after cutting the network connection photon apps
are
happy again and start as fast as with network.

I have no links of any kind to any other computers.

Augie

Friedhelm Schuetz
H.Kleinknecht & Co. GmbH

Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I fix
this?

Thanks

Augie

Hi

“Oleg Mityagin” <mityagin@rlt.ru> wrote in message
news:9a9bbi$2le$1@inn.qnx.com

We met this situation too often and our opinion is this could happen if
nameloc is running on each node of your network.
And when one node gets down, nameloc’s node polling makes the other nodes
slower.

How do you fix this? Do I have to stop nameloc? I only have nameloc
running on 2-3 nodes.

TIA

Augie

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > ???/??? ?
???
???: news:> 3AC7CF71.147134CB@kramer-smilko.com> …
The same sort of thing happens here when editing across the net.
Our node 2 is usuall missing, so I don’t know how fast vedit would
be with it present. Masking node 2 makes no noticeable difference.
Commands such as ls, on the other hand, are fast.

Maybe a vedit problem?

Richard

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take 3-6
seconds or more to get loaded. When everything is okay, Vedit Plus
brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop
on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I
fix
this?

Thanks

Augie

“Oleg Mityagin” <mityagin@rlt.ru> wrote in message
news:9a9bbi$2le$1@inn.qnx.com

We met this situation too often and our opinion is this could happen if
nameloc is running on each node of your network.
And when one node gets down, nameloc’s node polling makes the other nodes
slower.

I removed some nameloc’s running on other nodes. Now these nodes don’t have
licenses.

What should I do? How do I keep the usual performance, without losing my
licenses?

Please help.

TIA

Augie

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > ???/??? ?
???
???: news:> 3AC7CF71.147134CB@kramer-smilko.com> …
The same sort of thing happens here when editing across the net.
Our node 2 is usuall missing, so I don’t know how fast vedit would
be with it present. Masking node 2 makes no noticeable difference.
Commands such as ls, on the other hand, are fast.

Maybe a vedit problem?

Richard

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take 3-6
seconds or more to get loaded. When everything is okay, Vedit Plus
brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop
on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I
fix
this?

Thanks

Augie

use the -s and -e options of nameloc

Previously, Augie Henriques wrote in qdn.public.qnx4:

“Oleg Mityagin” <> mityagin@rlt.ru> > wrote in message
news:9a9bbi$2le$> 1@inn.qnx.com> …
We met this situation too often and our opinion is this could happen if
nameloc is running on each node of your network.
And when one node gets down, nameloc’s node polling makes the other nodes
slower.

I removed some nameloc’s running on other nodes. Now these nodes don’t have
licenses.

What should I do? How do I keep the usual performance, without losing my
licenses?

Please help.

TIA

Augie



“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > ???/??? ?
???
???: news:> 3AC7CF71.147134CB@kramer-smilko.com> …
The same sort of thing happens here when editing across the net.
Our node 2 is usuall missing, so I don’t know how fast vedit would
be with it present. Masking node 2 makes no noticeable difference.
Commands such as ls, on the other hand, are fast.

Maybe a vedit problem?

Richard

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take 3-6
seconds or more to get loaded. When everything is okay, Vedit Plus
brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop
on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I
fix
this?

Thanks

Augie

\

The only “bad” thing with this options is limitation of global names :frowning:
Thats why Sometimes you cant use -s & -e

“Ian Cannon” <ianc@tecom.com.au> ÓÏÏÂÝÉÌ/ÓÏÏÂÝÉÌÁ × ÎÏ×ÏÓÔÑÈ ÓÌÅÄÕÀÝÅÅ:
news:Voyager.010404085707.16663C@aresnode2.tecom.com.au…> use the -s
and -e options of nameloc

Previously, Augie Henriques wrote in qdn.public.qnx4:

“Oleg Mityagin” <> mityagin@rlt.ru> > wrote in message
news:9a9bbi$2le$> 1@inn.qnx.com> …
We met this situation too often and our opinion is this could happen
if
nameloc is running on each node of your network.
And when one node gets down, nameloc’s node polling makes the other
nodes
slower.

I removed some nameloc’s running on other nodes. Now these nodes don’t
have
licenses.

What should I do? How do I keep the usual performance, without losing
my
licenses?

Please help.

TIA

Augie



“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > ???/??? ?
???
???: news:> 3AC7CF71.147134CB@kramer-smilko.com> …
The same sort of thing happens here when editing across the net.
Our node 2 is usuall missing, so I don’t know how fast vedit would
be with it present. Masking node 2 makes no noticeable difference.
Commands such as ls, on the other hand, are fast.

Maybe a vedit problem?

Richard

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take
3-6
seconds or more to get loaded. When everything is okay, Vedit
Plus
brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance
drop
on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How
do I
fix
this?

Thanks

Augie


\

No - std path + usr/local/bin.

Bill Caroselli wrote:

Is anything in your PATH located on a filesystem that is no longer mounted?


Bill Caroselli - Sattel Global Networks
1-818-709-6201 ext 122

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC7CF71.147134CB@kramer-smilko.com> …
The same sort of thing happens here when editing across the net.
Our node 2 is usuall missing, so I don’t know how fast vedit would
be with it present. Masking node 2 makes no noticeable difference.
Commands such as ls, on the other hand, are fast.

Maybe a vedit problem?

Richard

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take 3-6
seconds or more to get loaded. When everything is okay, Vedit Plus
brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop
on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I
fix
this?

Thanks

Augie

Is anything in your PATH located on a filesystem that is no longer mounted?


Bill Caroselli - Sattel Global Networks
1-818-709-6201 ext 122



“Richard R. Kramer” <rrkramer@kramer-smilko.com> wrote in message
news:3AC7CF71.147134CB@kramer-smilko.com

The same sort of thing happens here when editing across the net.
Our node 2 is usuall missing, so I don’t know how fast vedit would
be with it present. Masking node 2 makes no noticeable difference.
Commands such as ls, on the other hand, are fast.

Maybe a vedit problem?

Richard

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take 3-6
seconds or more to get loaded. When everything is okay, Vedit Plus
brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop
on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I
fix
this?

Thanks

Augie

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take 3-6
seconds or more to get loaded. When everything is okay, Vedit Plus
brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance drop
on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How do I
fix
this?

Thanks

Augie

We investigated this a little bit one time, and came to the conclusion
that the culprit was either nameloc or Net.

Theory 1: When you shut down a node that is running nameloc, a local
nameloc will still attempt to synchronize with it periodically. This
synchronization takes a while, since the target node is down, so
nameloc essentially hangs for a little while. If nameloc is
_PPF_INFORMED, then it will effectively stop all processes from
starting or stopping on the node until it responds to any task death
message. A hung _PPF_INFORMED process will stop all task start/stops.
Eventually nameloc times out and answers the task death message, and
your processes load.

Theory 2: Net is involved in a similar kind of situation to what I
described above, directly.

Some points:

  • If you slay Net while you are in this “slow” state, your
    responsiveness will come back immediately.

  • If you wait long enough after shutting down the other node, then
    your responsiveness will suddenly come back. We think this is due
    to Net finally admitting that the other node is down.

  • You might be able to solve this problem by running automap, since it
    masks down nodes immediately in the netmap.

For what it’s worth, Augie, this is not a problem that only you are
having. It’s a fundamental network implementation bug in QNX 4. I
doubt very much that it will ever be fixed, or even looked at, by
QSSL. It has been there for as long as I can remember, and has been
reported many times. You will have to find a work-around. Automap is
probably your best avenue.

Cheers,
Andrew

“Andrew Thomas” <Andrew@cogent.ca> wrote in message
news:Voyager.010410131028.815126A@andrewhome.cogent.ca

Augie Henriques wrote:

“Richard R. Kramer” <> rrkramer@kramer-smilko.com> > wrote in message
news:> 3AC52FB3.A005B972@kramer-smilko.com> …
What aspect of performance got slower?

Compiling for one. When we edit files with Vedit Plus, they take
3-6
seconds or more to get loaded. When everything is okay, Vedit
Plus
brings
up the files instantaneously.

TIA

Augie



Augie Henriques wrote:

Hi.

I have a few QNX nodes running here (less than 10).

When I shutdown some/any of the nodes, I notice a performance
drop
on
some/all the other nodes running.

What’s this all about? Is the QNX networking acting up? How
do I
fix
this?

Thanks

Augie

We investigated this a little bit one time, and came to the conclusion
that the culprit was either nameloc or Net.

Theory 1: When you shut down a node that is running nameloc, a local
nameloc will still attempt to synchronize with it periodically. This
synchronization takes a while, since the target node is down, so
nameloc essentially hangs for a little while. If nameloc is
_PPF_INFORMED, then it will effectively stop all processes from
starting or stopping on the node until it responds to any task death
message. A hung _PPF_INFORMED process will stop all task start/stops.
Eventually nameloc times out and answers the task death message, and
your processes load.

Theory 2: Net is involved in a similar kind of situation to what I
described above, directly.

Some points:

  • If you slay Net while you are in this “slow” state, your
    responsiveness will come back immediately.

  • If you wait long enough after shutting down the other node, then
    your responsiveness will suddenly come back. We think this is due
    to Net finally admitting that the other node is down.

  • You might be able to solve this problem by running automap, since it
    masks down nodes immediately in the netmap.

For what it’s worth, Augie, this is not a problem that only you are
having. It’s a fundamental network implementation bug in QNX 4. I
doubt very much that it will ever be fixed, or even looked at, by
QSSL. It has been there for as long as I can remember, and has been
reported many times. You will have to find a work-around. Automap is
probably your best avenue.

Thanks. We see exactly what you mentioned. If anyone from QSSL is
listening, please reply? I would like to see if we can get a fix for this
problem?

Where can I get Automap? How much does it cost?

TIA

Augie

Cheers,
Andrew

Previously, Augie Henriques wrote in qdn.public.qnx4:

Where can I get Automap? How much does it cost?

Automap is written by Mario Charest. You can get it
either by going to www.zinformatic.com, or by sending
email to Mario at mcharest@zinformatic.com. The price
is $75 US for a 5-node license.

Frankly I think that automap should be installed on
any new QNX4 network. The price is low and the value
is high.

Cheers,
Andrew