Cluster Looses Its Virtual Ip's 1

[Teckman]

I Have Two Nodes In Cluster; Using MSCS. Nodes Are Connected To OutSide World Through A Switch. Which Further Connected To Our Router. After Every Week or So The Virtual Ip's Of Cluter Can't Be Accessed From The Diff. Network even in House. Physical address Can Be Accessed.
What Could Be The Problem????
Pl. Help
Thanks
Rajiv

 

[rmarston]

Sounds like a memory leak, do you have all the latest SP's
loded on both machines?

Which machine of the pair own's the IP address as far as the cluster admin is concerned?

If you move the Cluster IP address over to the other machine does it come back into life?

Is the IP address in a different range to any DHCP servers running?

Whats do the event logs say?

Rob

 

[Teckman]

Thanks, rmarston,
I will Look In For Memory leak. Both The Nodes Are Windows 2000 Advanced Server With Sp1 & Windows Critical Updates. installed. The M/c & Cluster Has static Ip Address. If I Move Cluster It Comes Online On Other Node. Event logs Dosen't Say Anything About This Problem. As Far As Cluster Working Is Concern It Works Fine Until It Looses The Ip Address. The Network Load Balancing Is Not Enabled. Is This Can Be A Router's Routing Table Problem or Somthing. I Am Still Confused. MultiCat or Unicast Problem.

 

[rmarston]

Sounds good, try moving the IP address to the other half of the cluster when you next loose the IP address! I assume any power saving modes have been turned off etc...

ROb

 

[Teckman]

rmarston,
what power saving modes u think could be problematic. i don't think it is the problem of the M/c Coz' we can access the ip's anytime on the same network.It becomes unavailable sometimes only from the other network. So i believe it has something to do with routers comming in the path.
what u think?

 

[rmarston]

If the network card were being put into low power it could be a problem. However as the address can be ping'd from the local LAN then it looks like this isn't the problem!

So what you are saying is...
{Get the * to line up vertically and the pic should appear}

LAN A LAN B *
Cluster A --+ *
| *
Cluster B --+----| Switch? |------ Computer B *
| Router? *
Computer A -+ *

Computer A can ping the Cluster IP address but Computer B can't?

Out of interest how does Computer B know the IP address of the Cluster network. Are you running WINS / DNS / LMHOSTS file across the different networks?

Rob

 

[Teckman]

hi,
let me explain u my set up once again.
I have two nodes with two lan cards in it one is for public network & other is for private network. Private network card is connected peer to peer with crossover cable. The public network cards of both computers are connected to a switch. The switch goes to some router ..I don’t know, some central circuits. Now there are four other computers are connected to the same switch that of cluster nodes. (local Lan). If u ping the ip’s from these M/c which are on the same switch & assigned their ips from the same range that of cluster nodes, ping all the time. But any M/c which is not on the same switch couldn’t get the ip’s sometimes, that is once in a month or once in couple of months or sometimes twice in a week…

 

[rmarston]

Ok, the topology is now clear, With just the issue of
how the post switch PC's know the IP addresses of
the cluster servers remaining.

Out of interest when the machines on the public LAN can't see the cluster server can they pin the cluster servers peer's {ie: the local PC's}

Rob

 

[Teckman]

No the private network is only to communicate between two nodes has ip as 192.168.x.x its not accessible at any time on lan or outside lan.

 

[rmarston]

No you misunderstand not the private circuit between the two cluster members. All referneces are to the public NIC's and associated switch

ROb

 

[Teckman]

yes. The Local m/c Can ping the cluster at any time.

 

[rmarston]

Ok, next time the cluster IP address fails try this...
{ping by both NAME and IP address }
{ie: ping clusterserver, ping 192.77.33.23 }

1) On a non local PC try to ping the cluster server
2) Run 'arp -a' on the non local machine and note its output
3) Run 'nbtstat -R' on the Non local Pc
4) Try to ping the cluster server again
5) Run 'tracert ServerIPAddress' on the non local PC and note the output

This should at least cut the size of the problem down

Rob

 

[cloudbuster]

Sounds like your Cluster servers are lossing the entry for the default gateway.
Next time it goes down, check IPCONFIG and make sure the default gateway entries are still there. Also try ROUTE PRINT and check the servers routing entries.

Also, when it goes down, try pinging a remote computer from your workstation...this should confirm correct router operation.
Chauncey

 

[AUTDOM]

Hi All,
I am having a problem with NT4 MSCS where the network cards one 1 server in the cluster is all red out .They show up in the cluster as unavailable.
This is the first time I am seeing this error since the box has been running for a year now.We had some down time and after rebooting ,the network cards became unavailable.
All network (3)cards can be pinged. 1 for wireless, 1 for wired network and 1 for the heartbeat. From server 1 you can ping all 3 IP's on the second cluster server.From server 2 you can ping all IP's on the server that has it's network cards unavailable in the cluster.
All resources are on server 2 and should not be like this when both servers are running.
Thanks,
Autdom...

 

[fgrasso]

The problem is not the cluster or the machine forming the cluster but the route.
When the cluster service switch an ip address on the other node use different nic with different phisical address, but the router cache also known the physial address related to the ip. When the cache is cleaned and the mac address is reread the problem solve.
To speed up the solution you have to manually clean the arp cache of the router when cluster switch or to set a lower refresh time of the cache (not suggested)