[Date Prev] [Date Index] [Date Next] [Thread Prev] [Thread Index] [Thread Next]

Automatic reinitialization, revisited

Steve Lammert slammert@panasas.com
Mon, 14 Oct 2002 22:13:00 -0700 (PDT)

Hi.  I've been a contented user of Conserver for several years, and 
currently have about 900 serial ports under management at three sites. 
Over time, we've used various Annex products, the Digi EtherLite, 
Cyclades TS-x000, and local devices, all with little or no trouble.

However, an attempt by one of our sites to try something new has 
generated major headaches for me.  They've purchased four 64-port 
Cyclades Ze cards and installed them in an x86 box running FreeBSD 4.3. 
  Cyclades provides a device driver (current version 2.1.1) which is 
then compiled into the BSD kernel.  After the subsequent reboot, one 
ends up with a slough of /dev/ttyCxx entries to talk to.

The fortunate thing is that the cards were installed in the machine 
which is our Conserver at that site.  Here's the unfortunate thing:  We 
can talk to these devices just fine using "tip" and similar tools, but 
have (so far) been unable to get Conserver to make use of them.

A simple conserver.cf entry of the form:


produces a string of error messages in /var/log/conserver of the form:

     conserver (6804): beta2-5-1: exit(3) [Mon Oct 14 21:51:39 2002]
     conserver (6804): beta2-5-1: automatic reinitialization [Mon Oct 14 
21:51:39 2002]

New messages appear once per second, per serial line.  Imagine the spew 
if you have 256 of these defined (:-(}

If this looks familiar, it may be because you recall a similar thread 
started by Malcolm Gibbs back in May.  His problems, though, seemed to 
center on reinitialization loops for lines which were actually down (not 
in use).  In my case, these lines are connected and active -- if I kill 
the conserver daemons (or just comment out the relevant lines and kill 
-HUP), then connect to the device using "tip", I can have a perfectly 
acceptable session.

As a workaround, we've configured Conserver to use "| tip <device>" in 
order to get things working.  But, gack, I really don't want to have 256 
"tip" sessions running all the time...

Is anyone out there using a similar setup (Cyclades Ze on FreeBSD?).  If 
so, was there any magic involved (e.g. settings in dev/cz/config.h 
before recompiling the kernel) in getting the devices to play nice with 
the Conserver?  or did it "just work" for you?

Steve L

steve lammert     software engineer   voice: +1-412-323-3500
slammert@panasas.com   panasas, inc     fax: +1-412-323-3511