On Thu, Jun 01, 2017 at 07:53:09PM +0000, Zbigniew Jędrzejewski-Szmek wrote:
On Thu, Jun 01, 2017 at 01:03:43PM -0500, Jason L Tibbitts III
> ZJ> That should not be a problem, but maybe this is tickling some bug in
> ZJ> the journal code?
> That's what it feels like. I might be able to hack something up where I
> force a journal rotation. Rotation of the journal doesn't invalidate a
> cursor, does it?
If the cursor points at a file which is gone, it becomes invalid. So yeah,
rotation of the journal can invalidate the cursor. In that case the code
should continue from the "beginning", i.e. the oldest available entries.
BTW., a debugging hint: if you need to generate a scenario where there
data is flowing in and journal files get rotated, it's easier to use
systemd-journal-remote and either feed it data from an existing
journal, or generate simple entries in python . This does not muck
up your real logs and is easier to control than a running
> ZJ> One suggestion would be to attach gdb to the hanging
script and look
> ZJ> the sd_journal object.
> I can do that. Is there any way to get at any useful debugging state
> from within python?
No, all the interesting information is in sd-journal code, even the C
part of the python wrapper is very thin.