[hobbit] still crashing

[Henrik Stoerner]

On Thu, Feb 08, 2007 at 04:00:47PM -0500, Rob Munsch wrote:
> I still have a constantly red-then-purple hobbitd_client on my hobbit 
> server.
> 
> It's gotten to the point where i have a cron job dropping the test 
> continuously.  I would appreciate any insight as to why this started 
> happening and what is causing it.
> 
> Core was generated by `hobbitd_client'.
> Program terminated with signal 6, Aborted.
> #0  0xffffe410 in __kernel_vsyscall ()

Unfortunately this doesn't give a clue about what actually happened,
except that it jumped to some wild address and crashed.

Could you add this line to hobbitd/hobbitd_client.c 
   dbgprintf("Client report from host %s\n", (hostname ? hostname : "<unknown>"));
around line 1754, just after the
    enum ostype_t os;
    namelist_t *hinfo = NULL;
lines. Then run "make" to rebuild hobbitd_client, copy the
hobbitd/hobbitd_client binary into ~hobbit/server/bin/ and edit
hobbitlaunch.cfg to include a "--debug" on the hobbitd_client command
(AFTER "hobbitd_client", ie at the end of the line).

hobbitd_client should restart automatically, and will be logging quite a
bit of data to the clientdata.log file, including the hosts that send it
data. This should let you figure out which host is sending the data that
triggers the crash, by comparing the time of the crash with the
timestamps in the logfile, or at least narrow it down.

Once you know which host it is, it would be interesting to see the
message this host sends. You can grab it from the "client data" link
for this host on the Hobbit web display. I'm obviously interested in
this message (please save it to a file instead of pasting it into an
e-mail), and also in the bb-hosts entry for the host, and any setup in
hobbit-clients.cfg.


Regards,
Henrik