|
Home > Archive > WebSphere Edge Server > January 2004 > What load = -1 does mean?
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
What load = -1 does mean?
|
|
| Nikolay Baranov 2004-01-19, 3:02 pm |
| Hi,
I had Edge Server installed on Win 2000 mashine.When I try to see load for
my servers (Current statistics tab for dispatcher advisor node), I always
see -1. What does it mean?
I think, this value indicate some existing troubles for my configuration.
Thanks,
N Baranov
| |
|
| Nikolay,
The Load of -1 under the Port column means that the advisor is unable to
open a connection with the server application. Either it's not
running, not responding to the cluster address, or there is a network
problem (routing, aliasing, etc0. I'd check the advisor log and maybe
start doing some network traces if the config looks correct.
Jeff
| |
| Nikolay Baranov 2004-01-19, 3:02 pm |
| Thanks for reply, Jeff
I check everything, but all seems to me is correct. I had IBM HTTP server
installed (port=80) on both servers Win2K machines. I can see it content
using URL like this http://server1:80 or http://server2:80 from dispatcher
machine. I can see its content via my claster adress also.
BUT advisor can not provide load info for me anyway. I checked http_80.log
file and discovered records with status = fail for both servers.
I find that the same problems can take place with some kind of NICs. I use
Netelligent 10T PCI UTP Controller. I can not find in its properties
something like Taskoffload property.
Have you got any info about this kind of NIC may to be the source of issue?
Are there to way detect (via logs, trace etc) that the problem is the NIC
used?
Thanks again,
N Baranov
"JLee" <leeja@NOSPAMus.ibm.com> wrote in message
news:3E91B030.2090609@NOSPAMus.ibm.com...quote:
> Nikolay,
>
> The Load of -1 under the Port column means that the advisor is unable to
> open a connection with the server application. Either it's not
> running, not responding to the cluster address, or there is a network
> problem (routing, aliasing, etc0. I'd check the advisor log and maybe
> start doing some network traces if the config looks correct.
>
> Jeff
>
| |
|
| Nikolay,
The other thing to check in the advisor log is how soon do you get the
FAIL messsages. If it's quick, then the server connection was refused,
but if it timed out then there could be a number of other problems. You
might want to do some traces on the ND machine and the servers to see
where the problem seems to be ... if you don't have the MS utility (I
forget what it's called), I'd recommend the Ethereal tool
(http://www.ethereal.com/). Once you isolate where the packets are
having problems, that would help determine what the problem is. If you
need assistance, you might want to open a PMR with the IBM L2 support
team and get direct phone support while doing this.
Jeff
| |
| Nikolay Baranov 2004-01-19, 3:02 pm |
| Thanks, Jeff
I discovered one interesting thing. I reconfigured my cluster to NAT support
(was Mac forwarding) and advisor seems work well. I can see positive and
realistic values for load. Therefore, NIC works properly. Probably, the
problem is in wrongly configuration for mac forwarding. but where exactly?
Have you any ideas about with this additional info?
N Baranov
"JLee" <leeja@NOSPAMus.ibm.com> wrote in message
news:3E92CE5D.3030709@NOSPAMus.ibm.com...quote:
> Nikolay,
>
> The other thing to check in the advisor log is how soon do you get the
> FAIL messsages. If it's quick, then the server connection was refused,
> but if it timed out then there could be a number of other problems. You
> might want to do some traces on the ND machine and the servers to see
> where the problem seems to be ... if you don't have the MS utility (I
> forget what it's called), I'd recommend the Ethereal tool
> (http://www.ethereal.com/). Once you isolate where the packets are
> having problems, that would help determine what the problem is. If you
> need assistance, you might want to open a PMR with the IBM L2 support
> team and get direct phone support while doing this.
>
> Jeff
>
| |
|
| Nikolay,
MAC means that the request when it arrives at the server machine will be
targeting the cluster IP address, same for the Windows advisor requests.
Linux and Solaris advisors target the server IP directly by default.
So if you could get a response from a client to the cluster IP, then the
advisor should have worked as well...
unless...was the client on the same local subnet? Or was it on another
subnet? Did you delete the loopback cluster route on the Windows
servers? If not, then local traffic (local clients and the advisors)
would not work but remote subnets would be fine.
In any case, to finish the above point, NAT changes it so that the
request reaches the server obviously targeting the server address. The
advisors change to target the server address as well and the cluster
loopback alias is no longer needed.
If the loopback cluster route was deleted, and changing to NAT fixed the
problem, then that would mean the server applications were not
responding for cluster IP requests or that the advisor request was not
reaching the servers (local routing error on the Windows ND machine).
Jeff
|
|
|
|
|