On Tue, Oct 27, 2009 at 04:25:57PM -0400, Anurag S. Maskey wrote: > > > Renee Danson Sommerfeld wrote: >> On Tue, Oct 27, 2009 at 03:49:17PM -0400, Anurag S. Maskey wrote: >> >>>> On Mon, Oct 26, 2009 at 09:41:37PM -0400, Anurag S. Maskey wrote: >>>> >>>>> This disabling of the datalink-management service is for bfu from >>>>> pre-dlmgmtd world to post-dlmgmtd world. Do we still need to >>>>> make this consideration or can we change the bfu script so that >>>>> datalink-management is not disabled anymore? >>>>> >>>> Unfortunately, we need to handle both cases. We can't just rip >>>> this out; we need to detect which case we're in, and adjust the >>>> behavior if it's the post-dlmgmtd to post-dlmgmtd scenario, while >>>> still doing the right thing (existing behavior) for the pre-dlmgmtd >>>> to post-dlmgmtd case. Ick. >>>> >>> Code review requested for >>> >>> 9947 NWAM causes network disruption when BFUing over NFS - >>> hanging BFU >>> http://defect.opensolaris.org/bz/show_bug.cgi?id=9947 >>> >>> webrev at >>> >>> >>> http://zhadum.east/export/ws/am223141/checkout-area/nwam1-fixes/webrev/ >>> >> >> This looks good. I'm just wondering if it would make sense to do >> the check for the presence of the service inside smf_import_service. >> It would mean you would always need to pass the FMRI as well as the >> manifest; and the check on line 542 would be of the 'svcs $FMRI' >> output rather than for the presence of the second parameter. Would >> that work? >> > Doing the svcs check in line 542 will work in this case, but it would > change the meaning of the second argument to smf_import_service() - > rather than it being used to specify if we want the service to be > enabled only after reboot, it would mean we want the service enabled > after reboot even if the manifest specifies "enabled=false".
Okay, sounds like there's a little more subtlety there than I realized. Seems like your solution is reasonable, then. -renee > Anurag > >>> The determination of post-dlmgmtd world is done by checking if "svcs >>> datalink-management" return 0 or not. >>> >>> Thanks, >>> Anurag >>> >>>>> -------- Original Message -------- >>>>> Subject: Re: disabling datalink-management in bfu >>>>> Date: Mon, 26 Oct 2009 13:16:06 -0700 >>>>> From: Cathy Zhou <Cathy.Zhou at Sun.COM> >>>>> To: Anurag S. Maskey <Anurag.Maskey at Sun.COM> >>>>> CC: Eric.Cheng at Sun.COM, clearview at sun.com >>>>> References: <4AE5EA41.5090006 at Sun.COM> >>>>> >>>>> I believe this is for bfu from pre-dlmgmtd to post-dlmgmtd, in >>>>> that case, there is no safe environment for dlmgmtd to run and >>>>> we expect it to be only run after reboot. >>>>> >>>>> I believe you can change the bfu script to skip that part if that >>>>> is not the case (if this is a bfu from post-dlmgmtd to >>>>> post-dlmgmtd). >>>>> >>>>> Thanks >>>>> - Cathy >>>>> >>>>>> Hi Cathy and Eric, >>>>>> >>>>>> bfu disables the datalink-management during the course of its >>>>>> execution. This has been presents NWAM phase 1 with problems >>>>>> as described in bug >>>>>> >>>>>> 9947 NWAM causes network disruption when BFUing over NFS >>>>>> - hanging BFU >>>>>> http://defect.opensolaris.org/bz/show_bug.cgi?id=9947 >>>>>> >>>>>> The summary is that datalink-management is disabled during the >>>>>> course of bfu. This disrupts the network as NWAM is told that >>>>>> links don't exist and it tears things down. >>>>>> >>>>>> I did some investigation and found the cause of the disabling. >>>>>> I wrote it out at >>>>>> http://defect.opensolaris.org/bz/show_bug.cgi?id=9947#c6. >>>>>> >>>>>> Is there a specific reason why the datalink-management service >>>>>> has to be disabled during bfu? If not, the fix would simply >>>>>> be to call smf_import_service() with just the dlmgmt.xml >>>>>> manifest. >>>>>> >>>>>> Thanks, >>>>>> Anurag >>>>>> >>>>>> >>>>> _______________________________________________ >>>>> nwam-dev mailing list >>>>> nwam-dev at opensolaris.org >>>>> http://mail.opensolaris.org/mailman/listinfo/nwam-dev >>>>> >>> _______________________________________________ >>> nwam-dev mailing list >>> nwam-dev at opensolaris.org >>> http://mail.opensolaris.org/mailman/listinfo/nwam-dev >>>
