| Author |
Message |
Johann J Hanekom
Guest
|
 Posted:
Mon Dec 27, 2004 4:46 pm Post subject:
Outlook loses connection to Exchange 2003 |
|
|
I have just completed a Windows 2003/Exchange 2003 installation on a
server with 1.5Gb memory - not sure if the rest of the specs are
relevant.
The startup is configured with the /3GB switch as per recommendation,
but none of the other performance settings have been implemented.
I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
The server has a single RAID 5 array split into two partitions. It
provides file sharing, and has AD and DNS running on it. As far as
possible the boot partition has been setup for the exclusive use of
the operating system and Exchange, with all created data such as
mailboxes, logs, files, etc hosted on a separate data partition.
All services are running normally, and there are no error entries in
the event viewer after the last re-boot. Internal mail can be sent and
received OK, and so can internet mail.
BUT
Outlook takes about 30 seconds to connect to the Exchange server and
open the mailbox (when it can connect) and sporadically is unable to
connect at all. When this happens, a re-boot of the client machine
usually restores the connection, but the system has not been running
long enough to claim this cure always works.
In any case, I don't think my 50 users are going to be very happy to
have to re-boot everytime they lose connection to Exchange.
Any idea where to start looking for the problem (and a solution)?
Thanks
|
|
| Back to top |
|
 |
Al Mulnick
Guest
|
| Posted:
Mon Dec 27, 2004 7:39 pm Post subject:
Re: Outlook loses connection to Exchange 2003 |
|
|
I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
Why aren't you starting there? If the user picks up their mail from the
POP3 gateway, why are you thinking that Exchange is involved in this issue?
Al
"Johann J Hanekom" <Johann.Hanekom@lycos.co.uk> wrote in message
news:1104144392.8bca4c442f2c58f2856b88b62f7b1de7@teranews...
| Quote: | I have just completed a Windows 2003/Exchange 2003 installation on a
server with 1.5Gb memory - not sure if the rest of the specs are
relevant.
The startup is configured with the /3GB switch as per recommendation,
but none of the other performance settings have been implemented.
I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
The server has a single RAID 5 array split into two partitions. It
provides file sharing, and has AD and DNS running on it. As far as
possible the boot partition has been setup for the exclusive use of
the operating system and Exchange, with all created data such as
mailboxes, logs, files, etc hosted on a separate data partition.
All services are running normally, and there are no error entries in
the event viewer after the last re-boot. Internal mail can be sent and
received OK, and so can internet mail.
BUT
Outlook takes about 30 seconds to connect to the Exchange server and
open the mailbox (when it can connect) and sporadically is unable to
connect at all. When this happens, a re-boot of the client machine
usually restores the connection, but the system has not been running
long enough to claim this cure always works.
In any case, I don't think my 50 users are going to be very happy to
have to re-boot everytime they lose connection to Exchange.
Any idea where to start looking for the problem (and a solution)?
Thanks |
|
|
| Back to top |
|
|
Bob Christian
Guest
|
| Posted:
Tue Dec 28, 2004 11:37 am Post subject:
Re: Outlook loses connection to Exchange 2003 |
|
|
I would have to agree with Al, but would like to add a few things.
Consider issues with DNS, authentication, user state, session connection
initiation, and other services (WINS, authentication, GPO) working over the
sattelite may take a while.
Remember that DNS does a query, after 1 second re-queries, and then queries
every 2 seconds after that for about 16 seconds. At about 7 seconds it will
roll to the second DNS server and perform the same function. It will, in
fact, switch from UDP to TCP (you see the word hesiod in the packet if you
sniff with netmon), and continue. Then come the WINS queries, broadcasts
(hope those are disabled), etc.
Satellites are complicated. If you can, switch to a network based upon a
landline. Where I live there are no less than 6 companies that are willing
to drop a T-1 into an office and do this for a very reasonable price.
Switching to an ASP that would host your server might even be an option.
It can take between 750ms to 6 seconds (6,000ms) to "squirt the bird", route
traffic, recieve a response, and have the traffic routed back to the
requesting client. That is just for a small packet, such as a ping, DNS
request, etc. I have seen these pings and ran both Netmon and Ethereal
captures on them. This is not my site, nor a constellation I have worked
with, but it gives you an idea of what you are facing:
http://www.ja.net/development/network_access/satellite/activity1/results.html
You may want to look at a utility called NeoTrace. It is a great tool and
reports response times along the path. It will even map this out based upon
location of the IP registration. It can't tell you are hitting a satellite,
but it will show the city of the provider where the IP. It will also show
where the latency is located. (Note: tracert does something similar for
free). Running it from a remote client (one across the satcom link) can
provide a lot of information.
You can have a 1 Megabit link or a 1 Gigabit sat link (they exist), but you
will still have the latency it takes for the radio signal to go from terra
firma (client) to the sattelite to terra firma (server - get processed) and
go back to the sattelite and back to terra firma (client). This,
unfortunately, is a function of physics and can't be fixed by buying more
bandwidth. Some providers overcome this by utilizing a 56k dial-up link for
requests (140ms on average for a small packet) and the response comes from
the sat.
Some of the latency can be fixed by switching constellations to a provider
that utilized lower-orbit sats. The Celestri (Motorola) and Teledesic
(private) networks were supposed to be the low orbit sats that took the net
to the sky (and reduced latency)....not sure what happened to them, but
Teledesic basically announced it would discontinue it's plans around 2 years
ago. Google it, their information is an interesting read. Irridium
provides contracted data services, from what I understand, and probably
serves the purposes these networks were to serve before the dot-com bust.
The Irridium sats are are 800km up versus 36,000km up for the telco sats.
GPS is ~11,000 nautical miles, or 20,372 km up there.
Consider this: it takes light (and radio waves (I am not a physicist))
about 6 seconds to reach the moon, 12 seconds for a round trip. It takes
GPS around 100ms to reach a reciever (GPS --> Reciever). Another thing to
consider is slant-range. You may be communicating with a satellite that is
11,000 nm in the sky, but as it comes over the horizon, it may be 17,000 nm
away from your location.
Bob
"Al Mulnick" <amulnick_No_SPAM@ncDOTrr.com> wrote in message
news:OTEv8lB7EHA.3076@TK2MSFTNGP15.phx.gbl...
| Quote: | I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
Why aren't you starting there? If the user picks up their mail from the
POP3 gateway, why are you thinking that Exchange is involved in this
issue?
Al
"Johann J Hanekom" <Johann.Hanekom@lycos.co.uk> wrote in message
news:1104144392.8bca4c442f2c58f2856b88b62f7b1de7@teranews...
I have just completed a Windows 2003/Exchange 2003 installation on a
server with 1.5Gb memory - not sure if the rest of the specs are
relevant.
The startup is configured with the /3GB switch as per recommendation,
but none of the other performance settings have been implemented.
I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
The server has a single RAID 5 array split into two partitions. It
provides file sharing, and has AD and DNS running on it. As far as
possible the boot partition has been setup for the exclusive use of
the operating system and Exchange, with all created data such as
mailboxes, logs, files, etc hosted on a separate data partition.
All services are running normally, and there are no error entries in
the event viewer after the last re-boot. Internal mail can be sent and
received OK, and so can internet mail.
BUT
Outlook takes about 30 seconds to connect to the Exchange server and
open the mailbox (when it can connect) and sporadically is unable to
connect at all. When this happens, a re-boot of the client machine
usually restores the connection, but the system has not been running
long enough to claim this cure always works.
In any case, I don't think my 50 users are going to be very happy to
have to re-boot everytime they lose connection to Exchange.
Any idea where to start looking for the problem (and a solution)?
Thanks
|
|
|
| Back to top |
|
|
Bob Christian
Guest
|
| Posted:
Tue Dec 28, 2004 11:50 am Post subject:
Re: Outlook loses connection to Exchange 2003 |
|
|
....and I know that I can't spell satellite properly.
"Bob Christian" <BobChristian@removethis.gmail.com> wrote in message
news:%23c6Br%23J7EHA.2032@tk2msftngp13.phx.gbl...
| Quote: | I would have to agree with Al, but would like to add a few things.
Consider issues with DNS, authentication, user state, session connection
initiation, and other services (WINS, authentication, GPO) working over
the
sattelite may take a while.
Remember that DNS does a query, after 1 second re-queries, and then
queries
every 2 seconds after that for about 16 seconds. At about 7 seconds it
will
roll to the second DNS server and perform the same function. It will, in
fact, switch from UDP to TCP (you see the word hesiod in the packet if you
sniff with netmon), and continue. Then come the WINS queries, broadcasts
(hope those are disabled), etc.
Satellites are complicated. If you can, switch to a network based upon a
landline. Where I live there are no less than 6 companies that are
willing
to drop a T-1 into an office and do this for a very reasonable price.
Switching to an ASP that would host your server might even be an option.
It can take between 750ms to 6 seconds (6,000ms) to "squirt the bird",
route
traffic, recieve a response, and have the traffic routed back to the
requesting client. That is just for a small packet, such as a ping, DNS
request, etc. I have seen these pings and ran both Netmon and Ethereal
captures on them. This is not my site, nor a constellation I have worked
with, but it gives you an idea of what you are facing:
http://www.ja.net/development/network_access/satellite/activity1/results.html
You may want to look at a utility called NeoTrace. It is a great tool and
reports response times along the path. It will even map this out based
upon
location of the IP registration. It can't tell you are hitting a
satellite,
but it will show the city of the provider where the IP. It will also show
where the latency is located. (Note: tracert does something similar for
free). Running it from a remote client (one across the satcom link) can
provide a lot of information.
You can have a 1 Megabit link or a 1 Gigabit sat link (they exist), but
you
will still have the latency it takes for the radio signal to go from terra
firma (client) to the sattelite to terra firma (server - get processed)
and
go back to the sattelite and back to terra firma (client). This,
unfortunately, is a function of physics and can't be fixed by buying more
bandwidth. Some providers overcome this by utilizing a 56k dial-up link
for
requests (140ms on average for a small packet) and the response comes from
the sat.
Some of the latency can be fixed by switching constellations to a provider
that utilized lower-orbit sats. The Celestri (Motorola) and Teledesic
(private) networks were supposed to be the low orbit sats that took the
net
to the sky (and reduced latency)....not sure what happened to them, but
Teledesic basically announced it would discontinue it's plans around 2
years
ago. Google it, their information is an interesting read. Irridium
provides contracted data services, from what I understand, and probably
serves the purposes these networks were to serve before the dot-com bust.
The Irridium sats are are 800km up versus 36,000km up for the telco sats.
GPS is ~11,000 nautical miles, or 20,372 km up there.
Consider this: it takes light (and radio waves (I am not a physicist))
about 6 seconds to reach the moon, 12 seconds for a round trip. It takes
GPS around 100ms to reach a reciever (GPS --> Reciever). Another thing to
consider is slant-range. You may be communicating with a satellite that
is
11,000 nm in the sky, but as it comes over the horizon, it may be 17,000
nm
away from your location.
Bob
"Al Mulnick" <amulnick_No_SPAM@ncDOTrr.com> wrote in message
news:OTEv8lB7EHA.3076@TK2MSFTNGP15.phx.gbl...
I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
Why aren't you starting there? If the user picks up their mail from the
POP3 gateway, why are you thinking that Exchange is involved in this
issue?
Al
"Johann J Hanekom" <Johann.Hanekom@lycos.co.uk> wrote in message
news:1104144392.8bca4c442f2c58f2856b88b62f7b1de7@teranews...
I have just completed a Windows 2003/Exchange 2003 installation on a
server with 1.5Gb memory - not sure if the rest of the specs are
relevant.
The startup is configured with the /3GB switch as per recommendation,
but none of the other performance settings have been implemented.
I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
The server has a single RAID 5 array split into two partitions. It
provides file sharing, and has AD and DNS running on it. As far as
possible the boot partition has been setup for the exclusive use of
the operating system and Exchange, with all created data such as
mailboxes, logs, files, etc hosted on a separate data partition.
All services are running normally, and there are no error entries in
the event viewer after the last re-boot. Internal mail can be sent and
received OK, and so can internet mail.
BUT
Outlook takes about 30 seconds to connect to the Exchange server and
open the mailbox (when it can connect) and sporadically is unable to
connect at all. When this happens, a re-boot of the client machine
usually restores the connection, but the system has not been running
long enough to claim this cure always works.
In any case, I don't think my 50 users are going to be very happy to
have to re-boot everytime they lose connection to Exchange.
Any idea where to start looking for the problem (and a solution)?
Thanks
|
|
|
| Back to top |
|
|
PES
Guest
|
| Posted:
Tue Dec 28, 2004 4:50 pm Post subject:
Re: Outlook loses connection to Exchange 2003 |
|
|
Bob Christian wrote:
| Quote: | ...and I know that I can't spell satellite properly.
|
In xp/2003 there is also a utility called pathping that does a lot of
what you can do with neotrace. It is non graphical but extremely useful.
| Quote: |
"Bob Christian" <BobChristian@removethis.gmail.com> wrote in message
news:%23c6Br%23J7EHA.2032@tk2msftngp13.phx.gbl...
I would have to agree with Al, but would like to add a few things.
Consider issues with DNS, authentication, user state, session connection
initiation, and other services (WINS, authentication, GPO) working over
the
sattelite may take a while.
Remember that DNS does a query, after 1 second re-queries, and then
queries
every 2 seconds after that for about 16 seconds. At about 7 seconds it
will
roll to the second DNS server and perform the same function. It will, in
fact, switch from UDP to TCP (you see the word hesiod in the packet if you
sniff with netmon), and continue. Then come the WINS queries, broadcasts
(hope those are disabled), etc.
Satellites are complicated. If you can, switch to a network based upon a
landline. Where I live there are no less than 6 companies that are
willing
to drop a T-1 into an office and do this for a very reasonable price.
Switching to an ASP that would host your server might even be an option.
It can take between 750ms to 6 seconds (6,000ms) to "squirt the bird",
route
traffic, recieve a response, and have the traffic routed back to the
requesting client. That is just for a small packet, such as a ping, DNS
request, etc. I have seen these pings and ran both Netmon and Ethereal
captures on them. This is not my site, nor a constellation I have worked
with, but it gives you an idea of what you are facing:
http://www.ja.net/development/network_access/satellite/activity1/results.html
You may want to look at a utility called NeoTrace. It is a great tool and
reports response times along the path. It will even map this out based
upon
location of the IP registration. It can't tell you are hitting a
satellite,
but it will show the city of the provider where the IP. It will also show
where the latency is located. (Note: tracert does something similar for
free). Running it from a remote client (one across the satcom link) can
provide a lot of information.
You can have a 1 Megabit link or a 1 Gigabit sat link (they exist), but
you
will still have the latency it takes for the radio signal to go from terra
firma (client) to the sattelite to terra firma (server - get processed)
and
go back to the sattelite and back to terra firma (client). This,
unfortunately, is a function of physics and can't be fixed by buying more
bandwidth. Some providers overcome this by utilizing a 56k dial-up link
for
requests (140ms on average for a small packet) and the response comes from
the sat.
Some of the latency can be fixed by switching constellations to a provider
that utilized lower-orbit sats. The Celestri (Motorola) and Teledesic
(private) networks were supposed to be the low orbit sats that took the
net
to the sky (and reduced latency)....not sure what happened to them, but
Teledesic basically announced it would discontinue it's plans around 2
years
ago. Google it, their information is an interesting read. Irridium
provides contracted data services, from what I understand, and probably
serves the purposes these networks were to serve before the dot-com bust.
The Irridium sats are are 800km up versus 36,000km up for the telco sats.
GPS is ~11,000 nautical miles, or 20,372 km up there.
Consider this: it takes light (and radio waves (I am not a physicist))
about 6 seconds to reach the moon, 12 seconds for a round trip. It takes
GPS around 100ms to reach a reciever (GPS --> Reciever). Another thing to
consider is slant-range. You may be communicating with a satellite that
is
11,000 nm in the sky, but as it comes over the horizon, it may be 17,000
nm
away from your location.
Bob
"Al Mulnick" <amulnick_No_SPAM@ncDOTrr.com> wrote in message
news:OTEv8lB7EHA.3076@TK2MSFTNGP15.phx.gbl...
I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
Why aren't you starting there? If the user picks up their mail from the
POP3 gateway, why are you thinking that Exchange is involved in this
issue?
Al
"Johann J Hanekom" <Johann.Hanekom@lycos.co.uk> wrote in message
news:1104144392.8bca4c442f2c58f2856b88b62f7b1de7@teranews...
I have just completed a Windows 2003/Exchange 2003 installation on a
server with 1.5Gb memory - not sure if the rest of the specs are
relevant.
The startup is configured with the /3GB switch as per recommendation,
but none of the other performance settings have been implemented.
I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
The server has a single RAID 5 array split into two partitions. It
provides file sharing, and has AD and DNS running on it. As far as
possible the boot partition has been setup for the exclusive use of
the operating system and Exchange, with all created data such as
mailboxes, logs, files, etc hosted on a separate data partition.
All services are running normally, and there are no error entries in
the event viewer after the last re-boot. Internal mail can be sent and
received OK, and so can internet mail.
BUT
Outlook takes about 30 seconds to connect to the Exchange server and
open the mailbox (when it can connect) and sporadically is unable to
connect at all. When this happens, a re-boot of the client machine
usually restores the connection, but the system has not been running
long enough to claim this cure always works.
In any case, I don't think my 50 users are going to be very happy to
have to re-boot everytime they lose connection to Exchange.
Any idea where to start looking for the problem (and a solution)?
Thanks
|
--
-------------------------
Paul Stewart
Lexnet Inc.
Email address is in ROT13 |
|
| Back to top |
|
|
Alan
Guest
|
| Posted:
Mon Jan 03, 2005 5:25 pm Post subject:
Re: Outlook loses connection to Exchange 2003 |
|
|
Check all the obvious infrastructure issues carefully.... DNS WINS ETC....
maybe check ure switch for network errors ( if u have a managed switch)
Is the server under hi IO load?? What speed hard disks r u running?
"Johann J Hanekom" <Johann.Hanekom@lycos.co.uk> wrote in message
news:1104144392.8bca4c442f2c58f2856b88b62f7b1de7@teranews...
| Quote: | I have just completed a Windows 2003/Exchange 2003 installation on a
server with 1.5Gb memory - not sure if the rest of the specs are
relevant.
The startup is configured with the /3GB switch as per recommendation,
but none of the other performance settings have been implemented.
I have a satellite broadband connection, with a gateway machine
between the server and the internet to provide NAT. The gateway
machine provides DHCP services, and hosts a local POP3 mail server
from which users pick up incoming e-mail.
The server has a single RAID 5 array split into two partitions. It
provides file sharing, and has AD and DNS running on it. As far as
possible the boot partition has been setup for the exclusive use of
the operating system and Exchange, with all created data such as
mailboxes, logs, files, etc hosted on a separate data partition.
All services are running normally, and there are no error entries in
the event viewer after the last re-boot. Internal mail can be sent and
received OK, and so can internet mail.
BUT
Outlook takes about 30 seconds to connect to the Exchange server and
open the mailbox (when it can connect) and sporadically is unable to
connect at all. When this happens, a re-boot of the client machine
usually restores the connection, but the system has not been running
long enough to claim this cure always works.
In any case, I don't think my 50 users are going to be very happy to
have to re-boot everytime they lose connection to Exchange.
Any idea where to start looking for the problem (and a solution)?
Thanks |
|
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in