Hi, Thanks much to Arkadiusz Miśkiewicz for pointing out that the watchdog daemon actually *does* write the "V". :-) I guess I was looking mostly at using the wd_keepalive daemon and it turns out that that daemon is the only one of the two that doesn't write the "V". I also noticed that the debian rules file residing where Arkadiusz pointed and has a comment indicating that the wd_keepalive daemon is still a work in progress and shouldn't be installed. So, it looks like the missing "V" may not be an issue right now.
Regarding the question of why anyone would want to stop the daemon, I guess there could be reasons why someone would want to stop it (temporarily). Maybe they would want to change the timeout/pretimeout/action/etc. settings during ipmi_watchdog module insertion time (rather than through the ioctl interface)? Another idea I'm wondering about is whether there might be an issue with the watchdog timer running (and being poked) while a firmware upgrade was being attempted or was underway? Anyway, thanks very much to you and Arkadiusz for the info and help, :-) Carol Hebert On Thu, 2006-09-28 at 20:45 -0500, Corey Minyard wrote: > The "V" character is standard for all watchdog drivers in Linux, so it > is a little surprising that the daemons don't support it. But not > terribly surprising. > > I guess the big question in my mind is why you would ever want to stop > the watchdog daemon for long enough to allow the watchdog to go off. It > seems you are opening a window for an unprotected failure. But your > best bet with this is to contact the mailing lists and authors > associated with those tools, as their function is outside of just IPMI. > > -Corey > > Carol Hebert wrote: > > Hi Corey, > > > > Thanks again to everyone who sent the excellent pointers to existing > > watchdog daemons. Two of the daemons at the ibiblio.org site (URL > > below) seemed to do everything I was hoping for (and more :-). The > > watchdog and wd_keepalive daemons seem to work very well with the > > ipmi_watchdog driver with the exception that they do not allow for a > > clean/expected shutdown of the timer upon close. Neither daemon writes > > the "V" character to the ipmi_watchdog driver to signal it to expect a > > close and to shut off the timer without triggering it. Unfortunately, > > removing the ipmi_watchdog module also does not shut off the timer. A > > raw ipmitool command can be sent subsequent to killing the daemon to > > shut off the timer (ipmitool -v raw 0x06 0x24 0x04 0x01 0x00 0x10 0x00 > > 0x0a) but it would probably be better to have the timer shut off > > automatically while performing a "service ipmi stop-watchdog". > > > > What do you think would be the best approach to having the timer shut > > off cleanly when the watchdog daemon is killed intentionally on an > > otherwise healthy system? > > > > Thanks much for your help, > > > > Carol Hebert > > > > On Mon, 2006-08-14 at 19:17 -0500, Corey Minyard wrote: > > > >> Doug Ambrisko wrote: > >> > >>> Dmitry Frolov writes: > >>> | * Carol Hebert <[EMAIL PROTECTED]> [12.08.2006 07:25]: > >>> | > I'm looking into what it would take to write a watchdog daemon and am > >>> | > trying to put together a list of requirements. I guess the first > >>> thing > >>> | > I need to know is whether anyone else is currently working on an ipmi > >>> | > watchdog daemon? If anyone on this mailing list is currently working > >>> on > >>> | > such a daemon, please let me know. > >>> | > > >>> | > The next question is (assuming an ipmi watchdog daemon isn't already > >>> in > >>> | > the works by someone): where would be the best place for such a daemon > >>> | > to reside? I think that including it with ipmitool might be the right > >>> | > place, since that's where ipmievd lives. Please let me know if you > >>> | > disagree and feel it belongs in the OpenIPMI library or elsewhere > >>> | > instead. > >>> | > > >>> | > I thought I'd also start discussing requirements and let those in this > >>> | > forum help decide what are the correct set of features/functionality. > >>> | > Please respond if you have any comments/opinions/questions on any of > >>> | > these points or if you have others you would like to have added to the > >>> | > list. > >>> | > >>> | Not that I know if anyone working on this, but here is FreeBSD's > >>> | watchdogd that can be used as a sample for watchdog daemon > >>> | functionality: > >>> | > >>> | > >>> http://www.freebsd.org/cgi/cvsweb.cgi/~checkout~/src/usr.sbin/watchdogd/watchdogd.c > >>> > >>> FWIW, the FreeBSD OpenIPMI like driver connects into the generic kernel > >>> watchdog so this just works under FreeBSD with watchdogd. So nothing > >>> has to be done except to use watchdogd. > >>> > >>> > >> The same is true for the Linux IPMI watchdog driver. It works just like > >> any other watchdog in the kernel except it has boot-time configuration > >> (and for newer 2.6 kernels, configuration through sysfs). This is > >> documented in the Documentation/IPMI.txt file. There is already a Linux > >> watchdog daemon at > >> http://www.ibiblio.org/pub/Linux/system/daemons/watchdog; it is fairly > >> thorough. > >> > >> -Corey > >> > ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Ipmitool-devel mailing list Ipmitool-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ipmitool-devel
