On Mon, Dec 12, 2011 at 6:25 PM, Nandisha H. M <nandi...@gmail.com> wrote: > As part of BMC health check we issue a cmd "ipmitool sunoem led get LOCATE"
I see. Thanks for explanation. > every 100 seconds (in timer function). We handled failure case like ipmitool > cmd return value is more than 0. But, we wondering if ipmitool cmd itself is Yes, checking return value is a common practice. That's good. > hung and never return back due to some issues (which you might explained). > Explanation later. > How important is to handle ILOM hung case in our code? > Well, it depends on how is the check executed. Is it executed periodically eg. from cron and doesn't it care whether previous run resp. ipmitool execution returned? In that case I can imagine server going down with N of ipmitool processes waiting for IPMI stack to reply. Something like "number of open FD is too big. Time to raise limits by # ulimit;" and such. Is it executed from web? Page probably won't finish loading and web server is likely to get stuck etc. In other words, yes, it is very important. But as I've said, I'm not familiar with Java and I don't know what your possibilities to handle this "issue" are. For example when using pexpect in Python, you can set timeout for execution resp. to match output. If not matched in given timeout, exception is raised. > >> But I did, I've tried it now and ipmitool got stucked and remains >> stucked and IPMI stack is stopped. > > Do you mean ILOM hung case reproduced? If so, please elaborate. Yes, I was able to reproduce it. I have access bellow and beyond ILOM's web interface, thus I can shut down(= turn off, kill processes etc.) IPMI stack completely. And ipmitool is going to hang and remain in this state for indefinitely. Shutting down IPMI stack on ILOM can simulate the case(which shouldn't happen, of course) of IPMI stack crash or being very slow to respond because of <whatever> reasons. I've simulated this by using service access to ILOM. If you can get beyond ILOM web interface, I can't say nor I can't give any ideas nor clues how to do so. Sorry. I'm sorry to say Harshad's patch is not going to help you even if I would implement it resp. patch ipmitool and would be able to give you a build. Still, even KCS interface can support time-out. Whether it should, whether it is a good idea, that's a question. I will start new thread about this one. I also am sorry you had to wait. But I wanted to take a closer look rather than quick replies and guessing which, I believe, resulted only in confusion. At the time, I was paying attention to other personal things and project. If you have more questions, need more explanation, just ask. Best regards, Z. > > Thanks in Advance, > Nandisha > > > On Mon, Dec 12, 2011 at 7:03 PM, Zdenek Styblik <zdenek.styb...@gmail.com> > wrote: >> >> On Mon, Dec 12, 2011 at 2:16 PM, Zdenek Styblik >> <zdenek.styb...@gmail.com> wrote: >> > On Mon, Dec 12, 2011 at 1:51 PM, Nandisha H. M <nandi...@gmail.com> >> > wrote: >> >> Thanks for the information Zdenek. >> >> >> >>> However, if you search this mailing list couple weeks back, you will >> >>> find patch from Harshad Prabhu which implements this feature. I think >> >>> you should be safe with this patch and ipmitool shouldn't get stuck in >> >>> case of IPMI gets AWOL. >> >> >> >> >> > >> > I gave it a thought, I've checked the patch ... well, actually subject >> > of the e-mail regarding the patch ... and it is related to >> > LAN/LAN-PLUS interfaces only. I'm sorry for confusion. >> > It still doesn't mean ipmitool can't get stuck when IPMI stack on ILOM >> > side is in some kind of trouble. >> > >> >> Hi Harshad, >> >> Could you please tell me which ilom patch/firmware has this feature? >> >> Should >> >> we need to pass any additional parm to ipmitool for timeout? >> >> >> >> I am currently using "ipmitool version 1.8.10.3". >> >> >> > >> > I'll bet it is from Hardware Management Pack, right? >> > >> >>> >> >>> In ILOM(I presume you have an Oracle/Sun server): >> >>> * Configuration -> System Management Access -> IPMI >> >>> * change state from Enabled to Disabled resp. uncheck the checkbox and >> >>> save >> >>> * try to get information via LAN -> ipmitool hangs. eg. % ipmitool -I >> >> >> >> >> >> I tried this setting in my ILOM web-tool. >> >> Configuration->SystemManagementAccess->IPMI and then unclick on Enable >> >> box >> >> (basically disable) and save. Even then I am able to execute below >> >> command >> >> without hang. >> >> >> > >> > Well, as I've said before ... Note, however, I was still able to query >> > ILOM via KCS interface. So you have to use LAN interface. >> > >> > Could you give out more info what the problem is or was? Because for >> > past ~ 8 months I haven't seen ipmitool to get stuck against ILOM >> > except the case I've described in previous e-mail. Of course one can't >> > rule out, perhaps and hopefully, extreme conditions on ILOM side. >> > However I doubt you could simulate those. >> > >> >> Kill IPMI stack in ILOM and ipmitool will get stucked even via KCS >> interface. ;) >> To question how to do it, how to kill IPMI stack in ILOM, I don't know. >> But I did, I've tried it now and ipmitool got stucked and remains >> stucked and IPMI stack is stopped. >> Funny enough, this didn't happen when SP being reset and I tried to >> query ILOM via KCS. >> >> Anyway, I will look at whole thing when I get back to work on Wednesday. >> I have caused enough confusion by guessing and don't want to cause any >> more. >> >> Z. > > ------------------------------------------------------------------------------ Cloud Computing - Latest Buzzword or a Glimpse of the Future? This paper surveys cloud computing today: What are the benefits? Why are businesses embracing it? What are its payoffs and pitfalls? http://www.accelacomm.com/jaw/sdnl/114/51425149/ _______________________________________________ Ipmitool-devel mailing list Ipmitool-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ipmitool-devel
