technical question for any networking people out there
#1
last night, in a fit of post-holiday enthusiasum, i decided to add a further two nic's to our main server here, taking the total to 4.
They are intel cards and are in a "load balancing" team i.e they appear to the OS as one card but obviously with potentially 4 x the throughput and fault tolerant as well.
Good idea you would think....
well...
this morning, when people were logging on, mostly 98 clients here, the logon script hung at the point where it maps the network drives. eventually it goes past that point but doesn't actually map the drives. If you look at the task manager, MPREXE is not responding. Now this mean that I have spent this morning running round and running the logon script manually which, on most machines has fixed the problem however I (and the users!!!) can do without this come monday morning !! (60 machines here.... ouch!!!)
The PDC is an NT 4.0 server with one Intel 10/100 card
The BDC (where the script runs from) is an NT 4.0 server with 2 x Intel 10/100 cards in a load balancing team
The server that the drives are mapped to is a Win2k server with (at present) 4 x 10/100 Intel cards in load balancing team.
The network is split across 3 x 10/100 Netgear switches.....
any ideas short of taking the cards out again ???
absolutely nothing i can find on technet re this...
They are intel cards and are in a "load balancing" team i.e they appear to the OS as one card but obviously with potentially 4 x the throughput and fault tolerant as well.
Good idea you would think....
well...
this morning, when people were logging on, mostly 98 clients here, the logon script hung at the point where it maps the network drives. eventually it goes past that point but doesn't actually map the drives. If you look at the task manager, MPREXE is not responding. Now this mean that I have spent this morning running round and running the logon script manually which, on most machines has fixed the problem however I (and the users!!!) can do without this come monday morning !! (60 machines here.... ouch!!!)
The PDC is an NT 4.0 server with one Intel 10/100 card
The BDC (where the script runs from) is an NT 4.0 server with 2 x Intel 10/100 cards in a load balancing team
The server that the drives are mapped to is a Win2k server with (at present) 4 x 10/100 Intel cards in load balancing team.
The network is split across 3 x 10/100 Netgear switches.....
any ideas short of taking the cards out again ???
absolutely nothing i can find on technet re this...
#2
Scooby Regular
Join Date: Nov 2001
Location: Leeds - It was 562.4bhp@28psi on Optimax, How much closer to 600 with race fuel and a bigger turbo?
Posts: 15,239
Likes: 0
Received 1 Like
on
1 Post
Do the switches support teams (spanning tree, I think??)
Try putting the cables Ie 1 out of each team into a different switch..
Do the servers have multiple Ip addresses on the same subnet? - not recommended..
Wouldnt bother with four cards for sixty users... well over the top...
fault tollerant team needs both links at the same speed and then fails over to the spare card... load balanced team requires clever switches normally... just use 2 cards in a fault tollerant team with 100 Full Duplex.. and See what happens.. I doubt you need more than that for 60 users... Why dont your scripts run from the pdc... they should run from either or have you done something stupid like stopped netlogon service on the pdc? Or have you not set your logonscript path to %logonserver% ?
Are you running netware or do you have the netware client installed as I seem to remember having this problem with netware
David
[Edited by David_Wallis - 7/19/2002 11:29:45 AM]
[Edited by David_Wallis - 7/19/2002 11:32:14 AM]
Try putting the cables Ie 1 out of each team into a different switch..
Do the servers have multiple Ip addresses on the same subnet? - not recommended..
Wouldnt bother with four cards for sixty users... well over the top...
fault tollerant team needs both links at the same speed and then fails over to the spare card... load balanced team requires clever switches normally... just use 2 cards in a fault tollerant team with 100 Full Duplex.. and See what happens.. I doubt you need more than that for 60 users... Why dont your scripts run from the pdc... they should run from either or have you done something stupid like stopped netlogon service on the pdc? Or have you not set your logonscript path to %logonserver% ?
Are you running netware or do you have the netware client installed as I seem to remember having this problem with netware
David
[Edited by David_Wallis - 7/19/2002 11:29:45 AM]
[Edited by David_Wallis - 7/19/2002 11:32:14 AM]
#4
the scripts always have run from the bdc for some reason.
but more to the point, they have until this morning run absolutely fine..
all i have done is add an extra 2 cards....
the teaming software that runs the cards gives the cards 1 ip address and makes them look like one card.
windows 2000 clients aren't affected... only 98... and nearly all the 98 machines allow manual running of the logon script after the machine has started up.... !!!!!
a little bit wierd... I'm going to try;
a. pulling the plug on the 2 extra cards so they ain't on the network anymore or
b. remove them from the team so they just sit there or
c. take them out of the server
but more to the point, they have until this morning run absolutely fine..
all i have done is add an extra 2 cards....
the teaming software that runs the cards gives the cards 1 ip address and makes them look like one card.
windows 2000 clients aren't affected... only 98... and nearly all the 98 machines allow manual running of the logon script after the machine has started up.... !!!!!
a little bit wierd... I'm going to try;
a. pulling the plug on the 2 extra cards so they ain't on the network anymore or
b. remove them from the team so they just sit there or
c. take them out of the server
#6
On Intel NIC S/W...
Adaptor Fault Tolerance and Adaptive Load Balancing don't need switch side support. Usually put all NICs into the same switch though.
Cisco FEC (Fast EtherChannel) and the Gigabit equiv. must have switch side support.
Adaptor Fault Tolerance and Adaptive Load Balancing don't need switch side support. Usually put all NICs into the same switch though.
Cisco FEC (Fast EtherChannel) and the Gigabit equiv. must have switch side support.
#7
hmmm...
well hopefully at the very least, when I remove the cards from the team, things will start working again and I will have to look at this when the office isn't full of people ..i.e a weekend
just regards the last post, Intel have said pretty much the same thing (and I did check this ages ago) put all the cards on the same switch. although to me that doesn't help the "fault tolerance" if the switch goes down !!
oh well
I will let you know what happens after 1:00
well hopefully at the very least, when I remove the cards from the team, things will start working again and I will have to look at this when the office isn't full of people ..i.e a weekend
just regards the last post, Intel have said pretty much the same thing (and I did check this ages ago) put all the cards on the same switch. although to me that doesn't help the "fault tolerance" if the switch goes down !!
oh well
I will let you know what happens after 1:00
Trending Topics
#10
Scooby Regular
Join Date: Nov 2001
Location: Leeds - It was 562.4bhp@28psi on Optimax, How much closer to 600 with race fuel and a bigger turbo?
Posts: 15,239
Likes: 0
Received 1 Like
on
1 Post
do a search for mpr exe....
Also set up the directory replicator service...
Connect to pdc\repl$\scripts put scripts in there set replicator to replicate to pdc and bdc... ensure logon script path doesnt have \\servername\netlogon\script.bat or whatever... just script.bat
ensure netlogon is started on pdc.. (should be or you'd have problems)
goto client machine and type echo %logonserver% to see what validated your request.
As with teaming... just meant it might need switch side support... wasnt sure... networking aint my specialised subject... however I do work with compaq servers and teaming cards (based on intel chipset) but never had major problems...
Do a search on google.com for mprexe as there are loads of things to check.
David
[Edited by David_Wallis - 7/19/2002 1:10:48 PM]
Also set up the directory replicator service...
Connect to pdc\repl$\scripts put scripts in there set replicator to replicate to pdc and bdc... ensure logon script path doesnt have \\servername\netlogon\script.bat or whatever... just script.bat
ensure netlogon is started on pdc.. (should be or you'd have problems)
goto client machine and type echo %logonserver% to see what validated your request.
As with teaming... just meant it might need switch side support... wasnt sure... networking aint my specialised subject... however I do work with compaq servers and teaming cards (based on intel chipset) but never had major problems...
Do a search on google.com for mprexe as there are loads of things to check.
David
[Edited by David_Wallis - 7/19/2002 1:10:48 PM]
#11
my hp switches do well with my 4 servers 1 with 4 intel cards load balenced, 2 more with 3 in and the last one with 2 in ok the last one is on 2 subnets but hey. I aint had any problems with load balancing with 170 users
#15
new drivers + pro-set software...
rebooted.. all seems fine now (it also runs a check to see if your switch config is ok as far as it is concerned and this came back fine)
strange how the other card worked fine though !!!
or maybe it wasn't... I suppose that could be something that just didn't manifest itself until there were three of the same card all having problems dealing with the network requests...
I'll have to see how it performs under full load
rebooted.. all seems fine now (it also runs a check to see if your switch config is ok as far as it is concerned and this came back fine)
strange how the other card worked fine though !!!
or maybe it wasn't... I suppose that could be something that just didn't manifest itself until there were three of the same card all having problems dealing with the network requests...
I'll have to see how it performs under full load
#16
Scooby Regular
Join Date: Nov 2001
Location: Leeds - It was 562.4bhp@28psi on Optimax, How much closer to 600 with race fuel and a bigger turbo?
Posts: 15,239
Likes: 0
Received 1 Like
on
1 Post
LMFAO at ChrisB
Until Not Freeze's Hell Over
Little puzzle for you..
Just had an interesting presentation from compaq though..
David
Until Not Freeze's Hell Over
Little puzzle for you..
Just had an interesting presentation from compaq though..
David
#18
Scooby Regular
Join Date: Nov 2001
Location: Leeds - It was 562.4bhp@28psi on Optimax, How much closer to 600 with race fuel and a bigger turbo?
Posts: 15,239
Likes: 0
Received 1 Like
on
1 Post
Blade servers
not supported unless you use their deployment software... my exact reply was (in this meeting) "Your having a f$cking laugh right?" er no... you can install manually, but we wont support you... "err why not..."... Dont Know...
I then asked if they were going to change the chassis without prior warning? erm no he said.. so I asked about the colour...
Dont like the sound of them myself, unless you are an isp or telco.. thought they might be ok for the citrix farm... but ill wait till we get some in..
Got an ML570 Quad Xeon to build soon... Ways a ton...
the rest is confidential and I obviously cant discuss
David
[Edited by David_Wallis - 7/19/2002 2:20:18 PM]
not supported unless you use their deployment software... my exact reply was (in this meeting) "Your having a f$cking laugh right?" er no... you can install manually, but we wont support you... "err why not..."... Dont Know...
I then asked if they were going to change the chassis without prior warning? erm no he said.. so I asked about the colour...
Dont like the sound of them myself, unless you are an isp or telco.. thought they might be ok for the citrix farm... but ill wait till we get some in..
Got an ML570 Quad Xeon to build soon... Ways a ton...
the rest is confidential and I obviously cant discuss
David
[Edited by David_Wallis - 7/19/2002 2:20:18 PM]
#24
ok... all seems ok now on the network front however, the reason I actually put the extra cards in was to address another issue I have come across since moving our data onto a win2k server;
file locking....
files that are in fact closed, are seen by the system as open and therefore cause problems i.e a program trys to write to a file, finds it still in use... waits....waits....waits...waits.... and all the time, the file ain't in use by anyone
so lets say for a program that reads a couple of access db files when it logs on, they should only be read very briefly at program startup and then released again except it seems the files are kept marked as in use which then causes problems when the next person tries to access them.....or logon to the program....
any ideas on this.
it used to work fine on an older slower server with S**t discs and only one NIC....
once again, this is something that doesn't seem to happen to anyone else !!!
file locking....
files that are in fact closed, are seen by the system as open and therefore cause problems i.e a program trys to write to a file, finds it still in use... waits....waits....waits...waits.... and all the time, the file ain't in use by anyone
so lets say for a program that reads a couple of access db files when it logs on, they should only be read very briefly at program startup and then released again except it seems the files are kept marked as in use which then causes problems when the next person tries to access them.....or logon to the program....
any ideas on this.
it used to work fine on an older slower server with S**t discs and only one NIC....
once again, this is something that doesn't seem to happen to anyone else !!!
#25
Scooby Regular
Join Date: Nov 2001
Location: Leeds - It was 562.4bhp@28psi on Optimax, How much closer to 600 with race fuel and a bigger turbo?
Posts: 15,239
Likes: 0
Received 1 Like
on
1 Post
poor programming?
dim conn as adodb etc..
prog..
conn = nothing
rs = nothing
Erm...
do a search on file locking timeout something like that...
probably a param under
HKLM\System\CurrentControlSet\Services\LanManServe r\Parameters
Or a Server Key.
dim conn as adodb etc..
prog..
conn = nothing
rs = nothing
Erm...
do a search on file locking timeout something like that...
probably a param under
HKLM\System\CurrentControlSet\Services\LanManServe r\Parameters
Or a Server Key.
#30
Ahh, not find it yet but maybe it wasn't file locking.
I vaguely remember the PC kept getting access denied to mapped shares. Log off / on and access would return. Let me double check.
I vaguely remember the PC kept getting access denied to mapped shares. Log off / on and access would return. Let me double check.