Re: [osol-discuss] milestone none really means nothing running at all really
Dennis Clarke [EMAIL PROTECTED] wrote: The boot -m milestone=none resulted in this : Booting to milestone none. Requesting System Maintenance Mode (See /lib/svc/share/README for more information.) Console login service(s) cannot run Root password for system maintenance (control-d to bypass): single-user privilege assigned to /dev/console. Entering System Maintenance Mode Nov 16 03:48:07 su: 'su root' succeeded for root on /dev/console -sh: /bin/i386: not found -sh: /usr/sbin/quota: not found -sh: /bin/cat: not found -sh: /bin/mail: not found /sbin/mount -m -r /usr . Jörg -- EMail:[EMAIL PROTECTED] (home) Jörg Schilling D-13353 Berlin [EMAIL PROTECTED](uni) [EMAIL PROTECTED](work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
[EMAIL PROTECTED] wrote: Yes. And we've long said you shouldn't be doign that :-) Show me the doc or white paper that says so and why. Read my Usenet postings :-) But the explanation is fairly simple: you cannot recover from a number of failures (corrupt vfstab, bad /dev* links for boot device) without having /usr mounted; you cannot mount /usr when those things happen. And you cannot recover if the same happens with a big / installation either. Jörg -- EMail:[EMAIL PROTECTED] (home) Jörg Schilling D-13353 Berlin [EMAIL PROTECTED](uni) [EMAIL PROTECTED](work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
But the explanation is fairly simple: you cannot recover from a number of failures (corrupt vfstab, bad /dev* links for boot device) without having /usr mounted; you cannot mount /usr when those things happen. And you cannot recover if the same happens with a big / installation either. You can, actually, rewire all of /dev if you are missing your /dev/dsk links. (Certain devices need to be present but the disk device need not be) Casper ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
Please explain why you believe this. the only difference I see in your case is that you would need to find the /devices entry to mount /usr so it makes sense to have find or a name completing shell in / When you boot from a device the node does not need to be present in /dev* in order for you to remoutn it r/w so you can fix it. The tools you need to create device nodes are all in /usr. If your disk device nodes are missing you can create them *if* you have /usr. Without /usr you can't save yourself. Casper ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
[EMAIL PROTECTED] wrote: Please explain why you believe this. the only difference I see in your case is that you would need to find the /devices entry to mount /usr so it makes sense to have find or a name completing shell in / When you boot from a device the node does not need to be present in /dev* in order for you to remoutn it r/w so you can fix it. This was true before /devices was on the devfs filesystem. Jörg -- EMail:[EMAIL PROTECTED] (home) Jörg Schilling D-13353 Berlin [EMAIL PROTECTED](uni) [EMAIL PROTECTED](work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
Joerg Schilling schrieb: [EMAIL PROTECTED] wrote: When you boot from a device the node does not need to be present in /dev* in order for you to remoutn it r/w so you can fix it. This was true before /devices was on the devfs filesystem. But having /usr ready helps a lot. It is no fun, cd'ing into /devices, doing an echo * just to find out where your /usr might be located. You may have to write a cat replacement in sh, cannot grep into files, etc. So with devfs it should be possible to recover with a separate /usr; OTOH having a unified / /usr eases recovery a lot. Daniel ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
Daniel Rock [EMAIL PROTECTED] wrote: Joerg Schilling schrieb: [EMAIL PROTECTED] wrote: When you boot from a device the node does not need to be present in /dev* in order for you to remoutn it r/w so you can fix it. This was true before /devices was on the devfs filesystem. But having /usr ready helps a lot. It is no fun, cd'ing into /devices, doing an echo * just to find out where your /usr might be located. I know that it definitely works as I did manage to manually mount the SchillIX CD /usr part when I was working on SchilliX-0.1 But I was happy because /opt/schily/sfind was available on the root fs ;-) You may have to write a cat replacement in sh, cannot grep into files, etc. There should be 'shcat' Jörg -- EMail:[EMAIL PROTECTED] (home) Jörg Schilling D-13353 Berlin [EMAIL PROTECTED](uni) [EMAIL PROTECTED](work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
On Wed, 2005-11-16 at 03:52, Dennis Clarke wrote: The boot -m milestone=none resulted in this : Booting to milestone none. Requesting System Maintenance Mode (See /lib/svc/share/README for more information.) Console login service(s) cannot run Root password for system maintenance (control-d to bypass): single-user privilege assigned to /dev/console. Entering System Maintenance Mode Nov 16 03:48:07 su: 'su root' succeeded for root on /dev/console -sh: /bin/i386: not found -sh: /usr/sbin/quota: not found -sh: /bin/cat: not found -sh: /bin/mail: not found # # df -ak df: not found # ls ls: not found # pwd /root That looks like you have split / and /usr, right ? In milestone none only the root filesystem is mounted read only. IMO this is REAL single user mode not that mamby pamby thing with all the local filesystems mounted and some basic networking running :-) -- Darren J Moffat ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
-sh: /bin/i386: not found -sh: /usr/sbin/quota: not found -sh: /bin/cat: not found -sh: /bin/mail: not found # # df -ak df: not found # ls ls: not found # pwd /root That looks like you have split / and /usr, right ? And we've long said you shouldn't be doign that :-) Not sure install follows that lead Casper ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
On Wed, 2005-11-16 at 15:01, [EMAIL PROTECTED] wrote: But I'm certain our best practices now are something like: / /var- for servers only /export - separate the users/data from the rest. I'd agree with that completely. In fact I'd actually say that on general purpose servers even having /var separate isn't useful. On a Kerberos KDC I've have /var/krb5 seperate but the rest of /var just part of /, similarly on a mail server using /var/mail that would be separate etc etc. I also tend to have /var/core and /var/crash separate as well but thats because I don't actually want live upgrade to copy the core and crash files over to the new boot environment, because they aren't relevant. Sometimes I just use dumpadm and coreadm to send them to a filesystem shared between all BEs. -- Darren J Moffat ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
[EMAIL PROTECTED] writes: - you run out of space more quickly because you dividing line will not be correct - hard recoverability issues become impossible with net/cdrom boot - the gain in stability is fairly minimal (a read-only /usr mount mostly increases maintenance costs. There's also the path of least resistance issue. The default installs don't include a separate /usr, which means that fewer people do this, and subtle bugs related to it are less likely to be caught. (I'm certainly not saying that our software isn't well-tested. Just that there's no good substitute for having large numbers of people using the system that way.) -- James Carlson, KISS Network[EMAIL PROTECTED] Sun Microsystems / 1 Network Drive 71.232W Vox +1 781 442 2084 MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677 ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
Darren J Moffat [EMAIL PROTECTED] writes: On Wed, 2005-11-16 at 15:01, [EMAIL PROTECTED] wrote: But I'm certain our best practices now are something like: / /var- for servers only /export - separate the users/data from the rest. I'd agree with that completely. In fact I'd actually say that on general purpose servers even having /var separate isn't useful. I separate / and /var because you can't mount / (incl. /usr) nosuid, but /var is fine with this option. You can even mount / nodevices nowadays, since device special files live in devfs as of S10 (and nosuid has been separated into nosetuid,nodevices; and old RFE of mine). Rainer -- - Rainer Orth, Faculty of Technology, Bielefeld University ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
DJM == Darren J Moffat Darren writes: DJM I also tend to have /var/core and /var/crash separate as well but DJM thats because I don't actually want live upgrade to copy the core and DJM crash files over to the new boot environment, because they aren't DJM relevant. DJM Sometimes I just use dumpadm and coreadm to send them to a filesystem DJM shared between all BEs. You can also symlink them off somewhere else. Matt -- Matt Simmons - [EMAIL PROTECTED] | Solaris Kernel - New York Ever consider what they must think of us? I mean, here we come back from a grocery store with the most amazing haul -- chicken, pork, half a cow. They must think we're the greatest hunters on earth! -- Anne Tyler ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
On Wed, 2005-11-16 at 16:21, Matthew Simmons wrote: DJM == Darren J Moffat Darren writes: DJM I also tend to have /var/core and /var/crash separate as well but DJM thats because I don't actually want live upgrade to copy the core and DJM crash files over to the new boot environment, because they aren't DJM relevant. DJM Sometimes I just use dumpadm and coreadm to send them to a filesystem DJM shared between all BEs. You can also symlink them off somewhere else. I use to do that but if using dumpadm and coreadm to actually put them else where seemed like a better solution to me - mainly because they have the functionality built in so why not use it. -- Darren J Moffat ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
DJM == Darren J Moffat Darren writes: DJM I use to do that but if using dumpadm and coreadm to actually put them DJM else where seemed like a better solution to me - mainly because they DJM have the functionality built in so why not use it. On the other hand, /var/crash is a well-known location, so there's some value in leaving something there that lets you get to the right place automagically. To each his own. DJM -- Darren J Moffat -- Matt Simmons - [EMAIL PROTECTED] | Solaris Kernel - New York The NASDAQ, properly understood, is nothing more than bingo for yuppies. - Rex Murphy, in an editorial on the CBC National News ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
On Wed, 2005-11-16 at 16:43, Matthew Simmons wrote: DJM == Darren J Moffat Darren writes: DJM I use to do that but if using dumpadm and coreadm to actually put them DJM else where seemed like a better solution to me - mainly because they DJM have the functionality built in so why not use it. On the other hand, /var/crash is a well-known location, so there's some value in leaving something there that lets you get to the right place automagically. To each his own. True. Which brings me to very minor grip about the default config of Solaris, it is /var/crash/hostname at the first time you booted which when using DHCP in the default Solaris config can change. Sure I can find out what it is from dumpadm. Why do we still have it as /var/crash/hostname rather than just /var/crash as the default given that dumpadm can be used to change it ? Is this something tied with how we built diskless clients on SunOS ? -- Darren J Moffat ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] milestone none really means nothing running at all really
DJM == Darren J Moffat Darren writes: DJM Which brings me to very minor grip about the default config of DJM Solaris, it is /var/crash/hostname at the first time you booted DJM which when using DHCP in the default Solaris config can change. Sure DJM I can find out what it is from dumpadm. DJM Why do we still have it as /var/crash/hostname rather than just DJM /var/crash as the default given that dumpadm can be used to change it? DJM Is this something tied with how we built diskless clients on SunOS ? I think it's historical. It *might* be tied to diskless clients. They've all got their own /var, but an admin could conceivably replace their /var/crash directories with symlinks to the mothership, but they could make the symlink hostname-specific just as easily. For the specific case you mentioned wrt DHCP, I believe there's an RFE open to add macro support to dumpadm (i.e. /var/crash/$HOST), but I don't have the ID handy. Matt -- Matt Simmons - [EMAIL PROTECTED] | Solaris Kernel - New York A report says the number of train derailments is up. What's the brand new slogan of Amtrak? We love to fly and it shows. - Alan Ray ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org