ATTN: Bill Smith

[Chris]

Bill,
There is a bunch of us posting with session timeout
errors, where the macs loss their shares intermittently
and the server event logs report:
_______________________
Source: MACSRV
Event ID: 12061

Session from user "<username>" was timed out and
disconnected by the server.
_______________________

Collin Barrett posted the same in November and referred
to a solution posted by Tony Sheppard in October. Tony's
post is no longer on the list. Another post claimed that
the fix was not a fix and that it is a server software bug.

Can you explain to us what is going on between our Macs
and servers and how to fix this problem?

Thanks,
Chris

 

[William M. Smith]

Bill,
There is a bunch of us posting with session timeout
errors, where the macs loss their shares intermittently
and the server event logs report:
_______________________
Source: MACSRV
Event ID: 12061

Session from user "<username>" was timed out and
disconnected by the server.
_______________________

Collin Barrett posted the same in November and referred
to a solution posted by Tony Sheppard in October. Tony's
post is no longer on the list. Another post claimed that
the fix was not a fix and that it is a server software bug.

Can you explain to us what is going on between our Macs
and servers and how to fix this problem?

Hi Chris!

I did a search on Google for '"Macs drop connection" and "Tony"' and found
the following:

===============================================================

This tend to be when a volume is not unmounted properly, IIRC.

For example ... if a volume is unmounted (dragged to trash) it is fine, but
if the machine crashes, or the machine is shutdown and fails to unmount
volumes / drives, or if it loses it's network connection, it takes some time
for SFM to respond. I vaguell recollect it being explained to me as SFM
doesn't want to let the session go in case it comes back after a short blip.

Since we have made sure the logout functions on our Macs definitely unmount
the volumes this has now dropped to a handful a week.

There is information about this in the Apple Knowledge Base somewhere. If I
can dig out the details i will post them here.

Tony Sheppard

.

I successfully tried Tony Sheppard's fix. I had a user
dismount, restart, then copy several files (she was
disconnecting with 1 MB files). Thank you for the solution.

Collin Barrett
===============================================================

Apparently, if you have a Mac that disconnects improperly due to a crash or
some other reason, you should restart your Mac, reconnect to the server and
then drag the server volume to the Trash to dismount properly. This should
tell the server to stop "holding on" to the old disconnected session.

Don't know if this is your solution, but hope this helps! bill

 

[Chris]

Bill,
First off thank you for your attention to this issue.
My drops occur when the machines are idle and when the
machines are in use. I get the following in the system log
when the drop occurs. I get exactly the same message if I
shutdown or restart the Macs without dragging mounted
volumes to the trash.
I still get the security event however I "unmount" or
logoff or get dropped.
I tried Tony's suggestion of unmount, restarting and
remounting and it has just failed. The failure was
predictable because we have been having to work around the
connection dropping issue on a particularly large 2GB job
when trying to write Postscript from Quark over the
network. We have not had an "idle machine" drop yet but it
is very intermittent and unpredictable at this time.
I will try to more thoroughly scour the Apple Knowledge
Base. It might be an Apple issue except that I get nearly
the identical behavior on one Win 2000 machine too, only
without the MACSVR system event.
I also have a question regarding, "tell the the server
to stop "holding on" to the old disconnected session."
because I am getting what appear to be the correct
sequence in the security events log however I disconnect
(or get disconnected).
Anything you can find out would be appreciated extremely
by me.


Thanks,
Chris Yates

 

[dnadeveloper]

For what it's worth, I get the problem too.
We just switched from old Novell servers to Win 2003 and I'm getting a huge
volume of complaints.
This problem is making me look quite bad. I will post anything I learn about
this problem and would appreciate anyone else doing the same.

Chris

 

[William M. Smith]

I tried Tony's suggestion of unmount, restarting and
remounting and it has just failed. The failure was
predictable because we have been having to work around the
connection dropping issue on a particularly large 2GB job
when trying to write Postscript from Quark over the
network. We have not had an "idle machine" drop yet but it
is very intermittent and unpredictable at this time.

Hi Chris!

You may already know this, but I'll throw it out for discussion. Windows
2003 still uses the old AFP 2.2 protocol, which is limited to Mac file names
of up to 31 characters and, I believe, is also limited to file sizes of less
than 2gb. I'd bet what you're seeing is a limit of the protocol. You can
test what I'm telling you by creating a file larger than 2gb on a Mac OS 9
system and trying to copy it to another Mac OS 9 system. It won't go.

You may need to look into something like ExtremeZ-IP from Group Logic, which
uses AFP 3.x. I believe you can download a 30-day demo for testing.

Just a thought. bill

 

[Chris]

Bill,
That's not it.
Today for example I have a file (and it supporting
documents) that worked fine on a Mac named g4-2 (OS 9)
yesterday. Today we changed one letter in the file
from 'a' to 'A' and today we get the 12061 MACSVR error in
the system events log when we try to write postscript from
g4-2.
To get the job out, we opened the exact same files from
the exact same volume on the server and successfully wrote
to the exact same location a postscript file from the Mac
named g4-1 (OS 9).
Results of this mornings test are totally inconclusive
and COMPLETELY BAFFLING:
4 times in a row the volumes were disconnected in the
process of writing the postscript files from the g4-2.
First, I printed the file to a laser from g4-2, to read
from the server but not "capture" the postscript file to
disk. This was successful. I thought, "this is
interesting."
Then, I wrote the postscript to the local disk on g4-2.
This was also successful. I am thinking, "Oh boy, I am
figuring something out here!"
Finally, I try writing the postscript back to the server
and it worked! And I thought, "@#%*!"
NOTHING has changed, except that it worked this time!
Continued searches of the Apple Knowledgebase have come
up with nothing.
Can you find out from Microsoft what the Error 12061
means?

Going bald,
Chris

BTW - "Idle" machine disconnects have continued to occur
on the OS X machine. In fact nothing was happening at all
on our whole system, it was diconnected after hours when
no one was here and no scheduled task, 9:05:08 PM, Error
12061.

 

[dnadeveloper]

Well, I think I found my problem with the Macs dropping but this may not be
anyone else's.
Basically I had network collisions because I was using multiple uplinks from
a switch that, unbeknown to me, did not support spanning tree.
Before fixing the multiple route problem, I was only pulling between 2 and
5% network utilization on the windows server now I pull just about 100% on
the 100megabit connections and 10% on the 10's.
Today is the first day but I'm sure this will take care of most of the
problems.


Chris

 

[Chris]

Chris,
I will check into this but all my Macs and my server are
on the same switch. Network utilization often approaches
100%. And this network was in place and running fine (with
one qualification to come) with everything the same except
that we replaced our old NT4 Server with a new 2003
Server. A quick test of the OS X machine copying a file
from one volume to another recorded utilization of 25 to
93% during the operation, other user were logged in but I
was the only one utilizing the server at that time.
The qualification is that my Mac OS X machine was
behaving the same if not dropping the connection even more
frequently as it is now. But only mine. I did not
troubleshoot it because it was ... because it was my
machine and like the cobbler's son goes barefoot...
The old NT Server was decommissioned and scavenged so I
doubt I could get to its old event logs to see what it had
to report.

Later,
Chris

 

[dnadeveloper]

I spoke to soon.
Network performance is good now but I do still drop some connections.
I suppose it was wishful thinking.
Back to the drawing board.
Chris

 

[William M. Smith]

Can you find out from Microsoft what the Error 12061
means?

Hi Chris!

This is from EventID.net:

Source: MACSRV

Type: Error

Description:
Session from user '<user name>' was timed out and disconnected by the
server. The IP address of the Macintosh workstation is in the data.

Ionut Marin (Last update 10/11/2003):
From a newsgroup post: "I had the same problem. It turned out to be a
faulty Ethernet switch/hub. Our NT workstations were getting disconnected as
well as our Macs, but this was harder to spot as NT reconnects in the
background. As soon as we replaced the switch, everything was Ok. Therefore,
my advice is to check out the infrastructure of your LAN: cabling, hubs,
switches, etc"

I think you said you've already checked hardware, but at least I thought I'd
point out the event description. Basically, the error code you're receiving
is just a reiteration of the message you're getting.

By any chance are the Macs you're using the Gigabit capable Macs? If so,
what happens if you install a 10/100 PCI card?

bill

 

[Chris]

Bill,
Built-in Gigabit Ethernet on the OS X machine, built-in
10/100 for the OS 9 machines and 10/100 card for the Win
2000. And a new one! a Win XP machine with a 10/100 Card
appeared in the Security Events Log this AM.
I am becoming discouraged and am suspicious that it is a
Server Software issue that I will never get an answer to
and that it will eventually be fixed in some patch or
update in a piggy back fashion with this specific issue
unannounced to the general public.
At least I hope it gets fixed.
I have posted to other groups and I am not alone but for
sure it does not seem to be happening universally, even in
predominantly Mac environments. If I could tell my
company, "if we buy this <switch, for instance>, then the
problem will go away." I would be very happy.
Interestingly, the Macs have had only one drop in the
last 24 hours since the "onslaught" of drops yesterday
morning. But we did get the first occurrence of the XP
machine.
Anything at all, guesses, suggested test, etc would be
greatly appreciated. However the intermittent nature makes
it much more difficult.
Is there any type of network monitoring software or
device that could check for hardware failures?


Later,
Chris

 

[William M. Smith]

Anything at all, guesses, suggested test, etc would be
greatly appreciated. However the intermittent nature makes
it much more difficult.
Is there any type of network monitoring software or
device that could check for hardware failures?

Hi Chris!

How about trying a new NIC in the server? Since you've identified the
problem is not file protocol specific (CIFS vs. AFP) then the we need to
work down the OSI layers (or DOD, whichever you prefer). I'm thinking you're
getting down to the physical layer. Could be as simple as a bad cable on the
server or going through multiple switches.

Hunches! All hunches!

Unfortunately, I'm not very familiar at all about protocol analyzers. Run
the Performance Monitor while you're copying some of those large files and
possibly the Task Manager as well. You may see something start peaking or
maxing out.

If you're running a major brand server such as Dell or Compaq, definitely
take advantage of their tools for monitoring the hardware. Your problem
could be specific to a network card or a particular set of drives.

Wish I could give you more! bill

 

[Chris]

Bill,
Interestingly enough, all the Macs stayed on all weekend
but the PCs kept reporting logon/logoff events in the
security log all day today. The PCs were pretty much not
in use if even started up this weekend.
As I said before the PCs "re-login" automatically when
the user attempts to open the share icon or network places.
A network card is realatively cheap so maybe that would
be a good plan.
Another forum recommended turning sleep off. They didn't
specify whether it was client or server. The server was
only set to turn off the monitor after 20 minutes but
never on the srives and system. So I'm going to leave that
as is and try the clients. The OS X machine was normally
disconnecting during idle times but the OS 9 machines were
normally in use. PCs are more difficult to tell since I
see the event logs after the fact and the users haven't
noticed anything because of the automated re-login.

Thanks,
Chris

 

[Glen Dahl]

I am having the same problems with macs falling off the
server. I contacted Asante to ask about spanning tree. If
you have macs on the network, you don't want spanning tree
as it will not coexist with appletalk.
The technician I spoke to suggested that I should look at
my network connections. Apparently, with cat 5 they are
finding more and more things that I did not know. He says
all cables, including cables from the computers to the
wall jacks should be at least 6 feet long. I know I have
some that are shorter. He also says that having a cable
run within three feet of a flourescent fixture can cause
problems especially for macs. I know that I have virtually
all of my lines that close to fixtures. It is almost
impossible not to.

This 12061 thing has been around for four years. I find
postings going back to 2000 with no answers. Microsoft
does not acknowledge this issue. I couldn't even get
anything from a search of their site.

I am going to check my wiring and see if I can fix it that
way.

 

[Tony Sheppard]

I am having the same problems with macs falling off the
server. I contacted Asante to ask about spanning tree. If
you have macs on the network, you don't want spanning tree
as it will not coexist with appletalk.
The technician I spoke to suggested that I should look at
my network connections. Apparently, with cat 5 they are
finding more and more things that I did not know. He says
all cables, including cables from the computers to the
wall jacks should be at least 6 feet long. I know I have
some that are shorter. He also says that having a cable
run within three feet of a flourescent fixture can cause
problems especially for macs. I know that I have virtually
all of my lines that close to fixtures. It is almost
impossible not to.

This 12061 thing has been around for four years. I find
postings going back to 2000 with no answers. Microsoft
does not acknowledge this issue. I couldn't even get
anything from a search of their site.

I am going to check my wiring and see if I can fix it that
way.

Just to stick my oar in here ... We seem to be having several different
issues here.

The dropping connection to an "idle" machine has indeed been round for years
and when we trialed ExtremeZ-IP it fixed it (and several other issues) but we
could never persuade management to let us buy it.

The problem with saving files working sometimes and not others we had 2 years
ago, and found it to be 9.1 being the problem. It then reappeared in 10.1 and
10.2.1 but quickly disappeared again.

When we changed our internal router we also had similar issues (originally an
ACC Colorado, but moved to a Cisco 2600) until we stuck a newer version of
the OS on.

Each time we had these problems we tried everything we had previously done
and then discovered it was something new. I only wish we could move onto
XServes and be done with it.

Tony Sheppard