Bug#793814: Workaround for this bug

[James Bottomley]

On Thu, 2015-10-15 at 01:06 +0200, Michael Biebl wrote:
> Am 15.10.2015 um 00:34 schrieb James Bottomley:
> > On Thu, 2015-10-15 at 00:30 +0200, Michael Biebl wrote:
> >> Am 15.10.2015 um 00:16 schrieb James Bottomley:
> >>> Since there doesn't seem to have been any progress on this from the
> >>> systemd side, and I'm really tired of my email users yelling at me
> >>> (justifiably since once the systemd-logind service dies, dovecot sasl
> >>> fails to work and they can't send email), this is the script I came up
> >>> with to alleviate the problem:
> >>
> >> So are you restarting dbus as well for some reason?
> > 
> > You can't really restart dbus in a systemd system...  However, the
> > problem seems to be the endpoint stops listening.  systemd-logind is
> > still running but strace -p shows it no longer responds when something
> > contacts the end point.  If you restart it, it will listen for a while
> > before stopping again.
> 
> If in your case you are not restarting dbus, can you isolate what's
> causing logind to stop responding on your system?
> I.e., can you provide us with steps how to reproduce the issue.

It happens naturally in a running system.  This one supplies
sparkleshare sync services for shared and backup directories, so there
are a huge number of micro sshd logins doing syncs, SASL via dovecot for
postfix and dovecot itself.  It's not based on LDAP, so everything goes
via the password file based authenticator.

The logs indicate that from reboot to first timeout took about 20 hours.
Thereafter, the timeouts seem to increase in frequency to about two or
three an hour.

James

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: This is a digitally signed message part
URL: <http://alioth-lists.debian.net/pipermail/pkg-systemd-maintainers/attachments/20151014/3cd468de/attachment-0002.sig>