| Remote processing of core dump [message #44432] |
Wed, 07 December 2011 11:51  |
Vasily Averin
Messages: 17
Registered: April 2008
|
Junior Member |
|
|
Hi Dave,
could you please advise is it possible to process core dump remotely?
Currently for investigation of kernel crashes on customer nodes I need either download coredump to local node or upload debug symbols to remote node. This may not be quite convenient for various reasons.
Is it possible to use gdbserver-like functionality with crash?
IMHO it would be great to start crash utility locally, with debug symbols but without core dump file, and then connect to gdbserver on remote node attached to core dump file.
Thank you,
Vasily Averin
|
|
|
|
| Re: Remote processing of core dump [message #44610 is a reply to message #44432] |
Wed, 07 December 2011 15:32  |
Dave Anderson
Messages: 1
Registered: December 2011
|
Junior Member |
|
|
----- Original Message -----
> Hi Dave,
>
> could you please advise is it possible to process core dump remotely?
No, not any more.
Actually I did implement that capability several years ago, where you
could build a remote "crashd" daemon that you could connect to in order
to access dumpfiles or /dev/mem on a remote system. But it languished due
to lack of interest and the preference/speed-superiority of using NFS,
and so it has long since been deprecated. The code still sits there in
the sources, in remote.c.
Dave
> Currently for investigation of kernel crashes on customer nodes I
> need either download coredump to local node or upload debug symbols
> to remote node. This may not be quite convenient for various
> reasons.
>
> Is it possible to use gdbserver-like functionality with crash?
>
> IMHO it would be great to start crash utility locally, with debug
> symbols but without core dump file, and then connect to gdbserver on
> remote node attached to core dump file.
>
> Thank you,
> Vasily Averin
|
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
