Re: [Condor-users] Trouble communicating through GCB servers

["Son, Sonny" <sson@xxxxxxxxxx>]

Sorry that you didn't get the answer first time when you tried GCB...

It seems that you have a misunderstanding on GCB (unless things have
been changed since I left Condor team). GCB supports private/firewalled
networks that allow outbound connections but block inbound connections.
NAT may be the best example that GCB can support: private nodes must be
able to initiate TCP connections or UDP sessions to the world.

In your setup, condor machines in network A must be able to talk to the
GCB servers of network A *and B*. Similarily, condor machines in network
B must be able to talk to the GCB servers of network B *and A*.


> -----Original Message-----
> From: m_kempenaar@xxxxxxxxx [mailto:m_kempenaar@xxxxxxxxx] 
> Sent: Monday, January 29, 2007 5:40 AM
> To: Condor-Users Mail List
> Subject: Re: [Condor-users] Trouble communicating through GCB servers
> 
> Hello,
> 
> Thanks for the replies everyone!
> 
> I've been experimenting some more with GCB, but still no success.
> I Now have the most basic setup I could think of to test GCB. 
> This setup consists of two private networks connected by a 
> router. One network connected to the WAN-side and the other 
> to the LAN-side, so both networks have different IP ranges.
> 
> See this picture for details:
> http://img95.imageshack.us/img95/9529/gcbprivatetoprivatenh6.png 
> 
> Each network has one GCB server and one Condor system (one 
> client in network A and a central manager in network B).
> Only the GCB-servers are told that there is a gateway 
> (router), so they can talk to each other perfectly. The 
> Condor systems do not know the route to each other or to the 
> GCB server from the other private network (as it will be in a 
> real-world situation where nodes cannot see each other directly).
> 
> I told both Condor systems to use the GCB-server within their 
> own private network (in condor_config.local) I believe this 
> works fine, since for example the MasterLog from 'condor 
> client #1' shows this line at startup: "DaemonCore: Command 
> Socket at <192.168.20.1:33778>" which is the address of the 
> GCB-server. But there isn't any traffic between both 
> GCB-servers. The central-manager is communicating through GCB 
> (while contacting it's own collector...) but the condor 
> client cannot contact the manager.
> 
> My guess is that it has something to do with the routing 
> tables for GCB.
> I've tried a number of settings, but it didn't help. It did 
> change some of the errors the condor client gave me. Examples 
> of the errors I get:
> 
> - "GCB: ERROR "GCB_TCP_connect: connection to the dest broker failed"
> errno 101: network is unreachable"
> - "GCB: [GCB_connect(12)]<192.168.50.2:9618>: direct connect 
> using _CB_do_connect failed"
> - "GCB: ERROR "GCB_TCP_connect: connection to the dest broker failed"
> errno 101: network is unreachable"
> 
> The gcb route table looked like this:
> 192.168.20.50/32  GCB   # public IP for GCB #2
> 192.168.20.1/32   GCB   # local GCB server
> 
> I use ethereal on GCB#2 to scan for any incoming traffic from 
> the .20.* network, but nothing comes through..
> 
> Can anyone help me constructing a gcb routing table?
> 
> Many thanks, and sorry for the long story once again...
> 
> Marcel Kempenaar
> 
> _______________________________________________
> Condor-users mailing list
> To unsubscribe, send a message to 
> condor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/condor-users
> 
> The archives can be found at either
> https://lists.cs.wisc.edu/archive/condor-users/
> http://www.opencondor.org/spaces/viewmailarchive.action?key=CONDOR
>