[PATCH] DON'T PANIC!
cs at samba.org
Tue Apr 23 17:26:27 UTC 2019
On Tue, Apr 23, 2019 at 01:41:51PM +0200, Ralph Böhme via
> Hi David,
> > Am 16.04.2019 um 14:33 schrieb David Disseldorp <ddiss at samba.org>:
> > Hi Ralph,
> > On Tue, 16 Apr 2019 14:02:47 +0200, Ralph Böhme wrote:
> >> On Mon, Apr 01, 2019 at 11:37:36AM +0200, Ralph Böhme via
> >> samba-technical wrote:
> >>> ping. :)
> >> 100% packet loss
> >> ping
> > Sorry about the packet loss, but my opinion hasn't changed here - I
> > still think panicking when the underlying filesystem has failed is
> > more desirable than exiting with 0.
> well, we log the underlying error so the logs contain all that's
> needed to analyse the problem. A SBT doesn't add anything helpful in
> this case.
> I guess we should generally only panick when facing an unrecoverable
> error in a Samba subsystem. In that case a SBT really helps, for every
> other case you just create work for the support team.
My 2¢: Samba should only panic due to an internal error on the Samba
side that needs debugging. If it is clear that the file system has a
problem, we don't need to fill the logs with Samba panics.
One example is that clustered file systems have problems (e.g. gpfs, but
probably others as well). The file system could lose quorum, disallow
any I/O on that node going forward. Having Samba panics in the logs
points then to the wrong component (Samba), instead of the actual
problem (quorum loss in the file system).
More information about the samba-technical