Keep in mind that part of the reason to have a BMC in the first place is to NOT impact the main CPU with so much management work, so doing management functions frequently like that should violate the primary goal of the customer's server: to do more production data throughput.
I've seen some customers do very frequent polling of every sensor, which takes more CPU and I/O bandwidth for management than is necessary or desirable. It is much more efficient to wait for IPMI events rather than poll the sensors so often. Let the BMC poll the sensors. Andy -----Original Message----- From: Al Chu [mailto:ch...@llnl.gov] Sent: Friday, March 26, 2010 7:16 PM To: Corey Minyard Cc: ipmitool-devel@lists.sourceforge.net; Andrew Wozniak Subject: Re: [Ipmitool-devel] ipmitool session intervals As Corey says, there *shouldn't* be a problem. I can recall off the top of my head at least three motherboards that had problems under heavy load. Heavy load here is not limited to IPMI but also to general network traffic (e.g. to many GARPs flying around on the network). Also, sensor readings can become affected if you do too much IPMI. For example, power readings or temperature readings are affected b/c you are doing too much IPMI on the board. Although you may consider IPMI monitoring just part of the system, and the consequence of it is just inherent in the sensor readings, I think most don't want that. "too much" is probably dependent on motherboard, but I know some people who monitor more than every 5 seconds, which I feel is "too much." Al On Fri, 2010-03-26 at 14:34 -0700, Corey Minyard wrote: > There *shouldn't* be an issue with this. However, BMCs do have bugs, so > if you are having problems, it's not a big surprise. > > -corey > > Andrew Wozniak wrote: > > Hi folks; > > > > First of all, thanks to everyone that supports/developes this > > essential tool for the IPMI community. > > > > I have one fundamental question. Are there any guidelines or > > limitations to how often ipmitool (over LAN) can be executed to insure > > robust interaction with a BMC? Since ipmitool uses a synchronous > > command/response protocol for each session, I would think not, as long > > as the BMC keeps up. > > > > Or is there a case where blasting the BMC with a _fast_ series of > > ipmitool operations will eventually cause the BMC to loose sync and > > freeze up? > > > > Regards, Drew > > > > > ------------------------------------------------------------------------ ------ > Download Intel® Parallel Studio Eval > Try the new software tools for yourself. Speed compiling, find bugs > proactively, and fine-tune applications for parallel performance. > See why Intel Parallel Studio got high marks during beta. > http://*p.sf.net/sfu/intel-sw-dev > _______________________________________________ > Ipmitool-devel mailing list > Ipmitool-devel@lists.sourceforge.net > https://*lists.sourceforge.net/lists/listinfo/ipmitool-devel > ------------------------------------------------------------------------ ------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ Ipmitool-devel mailing list Ipmitool-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ipmitool-devel ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ Ipmitool-devel mailing list Ipmitool-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ipmitool-devel
