[PATCH] ctdbd_conn: Generalise inaccurate error message

[Martin Schwenke]

Seeing:

  ctdb_read_packet failed: Cannot allocate memory
  [...,  0] ../source3/lib/ctdbd_conn.c:121(cluster_fatal)
  cluster fatal event: ctdbd died

The error is due to a memory allocation failure rather than ctdbd
dying.  However, the error message makes people wonder why ctdbd died.

Another alternative would be to wrap cluster_fatal() and have the
wrapper interpret the return value from ctdb_read_packet() to choose
from a set of more precise messages to pass to cluster_fatal().  For a
memory allocation it isn't strictly necessary to call cluster_fatal(),
but all is probably lost and it is still probably better to try to
exit cleanly as soon as possible instead of crashing somewhere.

Please review and maybe push...

I won't be offended to see a NACK and be asked to implement plan B as
mentioned above.  However, I figure that will lead to a certain amount
of bike-shedding without much gain.  The combination of both messages
already gives us all of the available information and I can't see a
case where we can definitely recover without resorting to
cluster_fatal().  Things are seriously wrong when we get to this
point....

peace & happiness,
martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-ctdbd_conn-Generalise-inaccurate-error-message.patch
Type: text/x-patch
Size: 2553 bytes
Desc: not available
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20181005/599be868/0001-ctdbd_conn-Generalise-inaccurate-error-message.bin>