Paul Legato <[EMAIL PROTECTED]> writes:
> No worries. I have a somewhat unusual setup networkwise. ... to say the least! > The controller is in a remote datacenter. The client is running in a > Debian Linux image running on my Macintosh laptop. By the way carob has been ported to darwinports. It would be great if you could give it a quick try and report! > So there are > actually several layers of firewalls involved.. but all pass ICMP > packets, and I can ping the controller from the client without problem. > The actual Carob connection is being forwarded through an SSH TCP > tunnel from the Debian image to the controller. (The client thus > thinks the controller is "localhost", which it is not.) We may have a problem here. The pinger is using UDP (same dest port as TCP). From what you say, carob is trying to ping localhost on UDP port 25322. Is there any chance it is going to get some answer...? > I don't know how Ruby's green threads interact with native threads in > an external library. Could it be that Ruby's threading system is > somehow interfering with the pinger threads in Carob? I would be surprised. Carob is using POSIX pthread_create() & co. Unless you did something _really_ convoluted at linking time, carob is going to use the native POSIX threads library of the system, whatever Ruby is doing above. I would be more concerned by the way your virtualization tool is scheduling linux threads. Which tool is it? _______________________________________________ Carob mailing list [email protected] https://forge.continuent.org/mailman/listinfo/carob
