Hi Mary, thank you very much for your help !
Jan mary ding wrote: > Jan: > > I am back from vacation and will give this a try for sparc and x86 > this week. Stay tuned. > > We had various x86 machines with 512 to 1GB memory for testing > purpose. I will need to see what I can find for sparc. > > > > jan damborsky wrote: >> Hi Mary, >> >> I have finally created reduced AI images based on 106 >> and built using IPS repositories created by David. >> >> They can be downloaded from following locations >> (NOTE: available only internally): >> x86 (size 157MB): >> /net/10.18.138.30/rpool/dc/ai_106_4166_final/media/automated_installer_image.iso >> >> >> >> Sparc (size 255MB): >> /net/tia.czech/data/dc/ai_106_4166_final/media/automated_installer_image_sparc.iso >> >> >> >> I have tested them on following machines: >> x86 HW: Ultra 20, 1GB RWM, no swap device created >> Sparc HW: T1000, 8GB RWM >> >> Installation was done from private 106 IPS repositories. >> >> When you have time, could I please ask you to give them a try ? >> If you would like to test on x86 machine with 1GB memory, >> please add following entry to the appropriate GRUB menu.lst >> file: >> >> # diff -u menu.lst._install_service_46501 >> menu.lst._install_service_46501_new >> --- menu.lst._install_service_46501 2008-11-28 13:08:05.478030411 >> +0100 >> +++ menu.lst._install_service_46501_new 2009-02-03 12:54:26.110704240 >> +0100 >> @@ -1,5 +1,6 @@ >> default=0 >> timeout=30 >> +min_mem64=1536 >> title OpenSolaris >> >> >> This change will be pushed as part of fix for bug 4166. >> >> It makes sure that the installation can succeed on system >> with 1GB memory without swap by running AI installation in >> 32bit mode. >> >> Thank you very much for your help, >> Jan >> >> >> >> mary ding wrote: >>> Jan and Dave: >>> >>> I am wondering whether we should use b106. b105 had a serious bug >>> with the e1000g driver: >>> >>> 6779610 e1000g fails to attach post-6713032 due to memory allocation >>> failures on some chipsets >>> >>> >>> Unfortunately a lot of our x86 test machines and server are using >>> e1000g, this will impact testing. >>> >>> >>> >>> jan damborsky wrote: >>>> Hi Mary, >>>> >>>> >>>> mary ding wrote: >>>>> Jan: >>>>> >>>>> Thanks for following up on this. I think we should leave the >>>>> SUNWpkgcmds alone. In the future, if we have the capability to >>>>> allow people to run custom finish scripts, I believe that pkgadd >>>>> command will be very handy do this. >>>> You and Dave convinced me - I will leave SUNWpkgcmds in AI image >>>> for time of being :-) >>>> >>>>> If you have AI image to test, let me know and I will be happy to >>>>> try them. >>>> I have just built AI image based on build 105. I have realized >>>> that microroot is bigger comparing to the one containing 101b >>>> bits. >>>> >>>> The difference is about 25MB (uncompressed) - I am going to >>>> investigate, why and if we could do anything about this. >>>> >>>> Once I have new image based on 105, I will let you know :-) >>>> >>>> Thank you very much for your help, >>>> Jan >>>> >>>>> >>>>> jan damborsky wrote: >>>>>> Hi Mary, >>>>>> >>>>>> >>>>>> mary ding wrote: >>>>>>> Dave and Jan: >>>>>>> >>>>>>> >>>>>>> I helped Jan to test out the fix with the reduced AI image and >>>>>>> it works. >>>>>>> >>>>>>> 1. However, I do notice that there are man pages in the >>>>>>> microroot and they do >>>>>>> occupy >>>>>>> some spaces. The man command is not even available in the >>>>>>> microroot, can these man packages be removed in the AI microroot: >>>>>> Based on later Karen's comment, I assume you mean >>>>>> that man pages are present in AI image/environment - >>>>>> I have verified that microroot itself doesn't contain >>>>>> directories you mention below. >>>>>> >>>>>> I have taken a closer look at what is present in AI image >>>>>> with respect to man pages and it seems that we can't >>>>>> get easily get rid of them at this moment. >>>>>> >>>>>> The reason is that they are spread across packages >>>>>> which deliver pieces we need in AI image. >>>>>> >>>>>> The good example might be SUNWopenssl package. >>>>>> We utilize only couple of libraries from it but >>>>>> it also delivers bunch of man pages and header >>>>>> files and the final size of that package is more >>>>>> than 10MB. >>>>>> >>>>>> I think that generic solution for this problem would >>>>>> be if we could apply filtering when installing >>>>>> IPS package into DC proto area - taking advantage >>>>>> of so called 'facets' would be the solution - please >>>>>> take a look at following thread, it gives the good >>>>>> introduction about facets/variants and how we could >>>>>> take advantage of them once they are implemented: >>>>>> >>>>>> http://mail.opensolaris.org/pipermail/pkg-discuss/2008-December/009092.html >>>>>> >>>>>> >>>>>> >>>>>> That said, I agree with you that there still might >>>>>> be candidates in AI image which are not needed >>>>>> and can be removed as a whole. >>>>>> >>>>>> I have created sorted list of packages we have >>>>>> in AI Distro Constructor manifest after Dave's >>>>>> reduction (list is attached). >>>>>> >>>>>> After taking a quick look, I can think at least >>>>>> following packages could be removed (but I have to >>>>>> verify): >>>>>> >>>>>> 10.66 MB SUNWsfinf GNU and open source info pages >>>>>> 6.33 MB SUNWgccruntime GCC Runtime libraries >>>>>> 4.54 MB SUNWgnu-coreutils coreutils - GNU core utilities >>>>>> 1.85 MB SUNWpkgcmds SVr4 packaging commands >>>>>> 1.59 MB SUNWgnutls GNU transport layer security library >>>>>> 1.36 MB SUNWpcre Perl-Compatible Regular Expressions >>>>>> 794.47 kB SUNWpython-cherrypy CherryPy >>>>>> >>>>>> Also, I plan to take a look at populated DC AI proto area, >>>>>> if there are some interesting packages installed which are >>>>>> not listed in AI manifest, but are pulled into the image >>>>>> in order to fulfill dependencies. >>>>>> >>>>>>> root at opensolaris:/usr# more /tmp/man >>>>>>> ./gnu/share/man >>>>>>> ./has/man >>>>>>> ./sfw/man >>>>>>> ./sfw/share/man >>>>>>> ./share/man >>>>>>> root at opensolaris:/usr# du -skh /usr/gnu/share/man >>>>>>> 235K /usr/gnu/share/man >>>>>>> root at opensolaris:/usr# du -skh /usr/has/man >>>>>>> 108K /usr/has/man >>>>>>> root at opensolaris:/usr# du -skh /usr/sfw/man >>>>>>> 512 /usr/sfw/man >>>>>>> root at opensolaris:/usr# du -skh /usr/sfw/share/man >>>>>>> 2.8M /usr/sfw/share/man >>>>>>> root at opensolaris:/usr# du -skh /usr/share/man >>>>>>> 6.1M /usr/share/man >>>>>>> root at opensolaris:/usr# >>>>>>> >>>>>>> >>>>>>> >>>>>>> 2. For Live CD, the man pages also occupy some space and should >>>>>>> they be removed also ??? In fact, just /usr/share/man alone had >>>>>>> already occupy 56 MB and this will definitely help the low >>>>>>> memory 512 MB >>>>>>> system doing livecd install. >>>>>> As far as LiveCD case is concerned, I think man pages >>>>>> are mandatory there, since LiveCD represent working >>>>>> Solaris environment and we need to provide potential >>>>>> new user with as much help/docummentation/man pages >>>>>> as possible in order to simplify the Solaris adoption. >>>>>> >>>>>> Thank you, >>>>>> Jan >>>>>> >>>>>> >>>>>> ------------------------------------------------------------------------ >>>>>> >>>>>> >>>>>> _______________________________________________ >>>>>> caiman-discuss mailing list >>>>>> caiman-discuss at opensolaris.org >>>>>> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss >>>> _______________________________________________ >>>> caiman-discuss mailing list >>>> caiman-discuss at opensolaris.org >>>> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss >> >> _______________________________________________ >> caiman-discuss mailing list >> caiman-discuss at opensolaris.org >> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss >
