CUPS in HA configurations

[Jason Czerak]

After some toying around adding printers things get weird.  If anything id have to shutdown one node, make my changes, bring up the second node if i'm to share /etc/cups.

Still working though the other question I have here my self.


> cupsd is not designed to use a shared configuration or spool directory, =
> and "Bad Things (tm)" will happen if you do so.
>
> Sharing /usr/share/cups is fine, of course, but don't expect /etc/cups =
> or /var/*/cups to be safe.
>
> On Dec 16, 2010, at 6:07 PM, Jason Czerak wrote:
>
> > I'm configuring an HA pair on 1.3.7 on RHEL 5.3. I'll be doing an =
> upgrade on things to RHEL 5.5 in the spring. I have a few questions that =
> you folks may be quick to answer before I waste paper and time. :)
> >=20
> > I'm experimenting with placing /etc/cups on an NFS mount (netapp). So =
> far the HA pair work well. if you fail the server or pause the queues =
> print jobs work as expected.
> >=20
> > /var/spool/cups and /var/cache/cups are local to each node at this =
> time.  I'm curious if I can share these locations as well? the job ID =
> file is the first hurdle. If I do share /var/cache/cups I'll ensure it's =
> mounted with actimeo=3D0. I'm curious if I'm to expect a race or locking =
> condition if both nodes in the HA pair try to update this file with the =
> next ID?
> >=20
> > The idea here is to save the spooled data. I may have a requirement =
> that I keep printed data for a period of time so if a reprint is needed =
> the spooler can take care of that. This way I can have one spot for the =
> data, not two
> >=20
> > /var/cache/ppds.dat. I've noticed that the second node doesn't have =
> this file and is working well. I have yet to add a printer while both =
> are sharing the /etc/cups data. What issues might I expect with this =
> data file out of sync?
> >=20
> > The main idea here with sharing /etc/cups is maintenance of the =
> queue's. I'm going to end up with near 100. I'd like to ensure both =
> servers are identical so there's not oddities.
> >=20
> > Second big question is what happens to a large print job that the cups =
> server fails mid spool? If client "C" is spooling to server "A" and =
> server "A" disappears (crashes, hardwareissue, network issue, etc) mid =
> spool will client "C" restart the spool to server "B"? or fail?
> >=20
> >=20
> >=20
> > _______________________________________________
> > cups mailing list
> > cups at easysw.com
> > http://lists.easysw.com/mailman/listinfo/cups
>
> ________________________________________________________________________
> Michael Sweet, Senior Printing System Engineer, PWG Chair
>