Date: Wed, 19 Nov 2014 23:15:05 +0100 From: Andreas Nilsson <andrnils@gmail.com> To: John Baldwin <jhb@freebsd.org> Cc: Daniel O'Connor <doconnor@gsoft.com.au>, FreeBSD Stable Mailing List <freebsd-stable@freebsd.org>, Dmitry Morozovsky <marck@rinet.ru> Subject: Re: SuperMicro IPMI/SOL and ipmitool troubles Message-ID: <CAPS9%2BSuP%2Bi6W5VHkx7gaec=QqKimEwGd1t30sdz8zN6WVcCGiQ@mail.gmail.com> In-Reply-To: <3870083.h75yLjv6As@ralph.baldwin.cx> References: <alpine.BSF.2.00.1411112022500.25949@woozle.rinet.ru> <CAPS9%2BSujGtrQ%2Bw=5jCBqV1vv_ZFZ5qkhjfGQUwc5n0sfjPCvnQ@mail.gmail.com> <CAPS9%2BSuCLE9kVPA-RNMHfziJYXAL1ivkV-B8wa64athYdDb38A@mail.gmail.com> <3870083.h75yLjv6As@ralph.baldwin.cx>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Nov 19, 2014 at 8:05 PM, John Baldwin <jhb@freebsd.org> wrote: > On Wednesday, November 19, 2014 05:02:49 PM Andreas Nilsson wrote: > > On Wed, Nov 19, 2014 at 3:28 PM, Andreas Nilsson <andrnils@gmail.com> > wrote: > > > On Fri, Nov 14, 2014 at 7:30 PM, Dmitry Morozovsky <marck@rinet.ru> > wrote: > > >> Daniel, > > >> > > >> nice to see you here too ;) > > >> > > >> On Fri, 14 Nov 2014, Daniel O'Connor wrote: > > >> > On 12 Nov 2014, at 19:43, Andreas Nilsson <andrnils@gmail.com> > wrote: > > >> > > unclear is the word for it :) And thanks for looking into this. > > >> > > >> ipmi/ilo is > > >> > > >> > > important on a server os. > > >> > > >> > > I found a reference to it in a ML post: > > >> > http://lists.freebsd.org/pipermail/freebsd-stable/2013-February/072464.ht > > >> ml > > >> > > >> > I started that thread :) > > >> > I did get it working on the hardware I was using (Supermicro X9SCL-F > > >> > > >> and X8SIL-F) > > >> > > >> > I used the following BIOS settings > > >> > > > >> > ? Remote Access - Enabled > > >> > ? Serial Port Number - COM3 > > >> > ? Serial Port Mode - 115200, 8, n, 1 > > >> > ? Flow Control - Hardware > > >> > ? Redirection After BIOS POST - Always > > >> > ? Terminal Type - VT100 > > >> > ? VT-UTF8 Combo Key Support - Disabled > > >> > ? Sredir Memory Display Delay - No Delay > > >> > > > >> > And the following in loader.conf > > >> > # Give preference to VGA console > > >> > console="vidconsole,comconsole" > > >> > # Uncomment below and comment above to give serial console > preference > > >> > #console="comconsole,vidconsole" > > >> > comconsole_speed="115200" > > >> > boot_multicons="YES" > > >> > hint.uart.0.flags="0x0" > > >> > hint.uart.2.at="isa" > > >> > hint.uart.2.port="0x3E8" > > >> > hint.uart.2.flags="0x30" > > >> > > > >> > And this in /etc/ttys > > >> > # IPMI console > > >> > # Note: The Java console viewer doesn't seem to be very smart as it > > >> > > >> doesn't > > >> > > >> > # properly support VT100 > > >> > cuau2 "/usr/libexec/getty 3wire.115200" vt100 on secure > > >> > > > >> > I could then access it using ipmitool like so > > >> > ipmitool -H remoteip -U ADMIN -I lanplus sol activate > > >> > [login] > > >> > export TERM=xterm > > >> > > > >> > Note that I wanted vidconsole by default because mostly the systems > > >> > > >> were used by people local to them, however we could break into the > loader > > >> and type 'set console=comconsole,vidconsole? and then get everything > over > > >> the serial console for remote trouble shooting. > > >> > > >> > You may also wish to check the IPMI configuration via the web > interface > > >> > > >> - by default it will failover to port 0 and it has terrible default > > >> passwords. I changed the passwords and forced it to use the dedicated > > >> IPMI > > >> port even if nothing was connected to it. > > >> > > >> Well, I'm almost done with most of our SM server, even concentrated > > >> console on > > >> our console server with such a simple config: > > >> > > >> ---- 8< ---- > > >> # ipmi/sol console template > > >> default ipmi { > > >> > > >> master localhost; > > >> type exec; > > >> exec /usr/local/bin/ipmitool -f /usr/local/etc/ipmi-pass -U > > >> > > >> root -I lanplus -H %.int sol activate; > > >> > > >> execsubst %=cs; > > >> #idletimeout 6h; > > >> > > >> break 0 { string "~B"; } > > >> > > >> } > > >> > > >> console gwn1 { include ipmi; } > > >> console gwn2 { include ipmi; } > > >> console gwn3 { include ipmi; } > > >> console gwn4 { include ipmi; } > > >> console gwn5 { include ipmi; } > > >> console gwn6 { include ipmi; } > > >> console gwn7 { include ipmi; } > > >> console gwn8 { include ipmi; } > > >> > > >> console gwc2 { include ipmi; } > > >> ---- 8< ---- > > >> > > >> This has console logging (including possible panics) as a surplus > > >> > > >> -- > > >> Sincerely, > > >> D.Marck [DM5020, MCK-RIPE, > DM3-RIPN] > > >> [ FreeBSD committer: marck@FreeBSD.org > ] > > >> > ------------------------------------------------------------------------ > > >> *** Dmitry Morozovsky --- D.Marck --- Wild Woozle --- marck@rinet.ru > *** > > >> > ------------------------------------------------------------------------ > > > > > > Hello again, > > > > > > Searching on hw.uart.console, I found: > > > > http://lists.freebsd.org/pipermail/svn-src-head/2013-February/044641.html > > > , a very enlightening thread. > > > > > > Basically: "ohh, you want to use something other than COM1 and tried to > > > get away with just changing hint.uart stuff, which has worked for a > while, > > > ha, no way..." No heads up, nothing. > > > > > > Sorry to say jhb@ but is not a rare case. It is if not the default, a > > > very common setup on every HP server with iLO, and it holds for most > all > > > OOB style serial emulation I have ever had the (dis)pleasure of working > > > with. > > This was done _specifically_ so you could use non-COM1 for both loader and > kernel with one thing to change. That is, you don't use hint.uart.X.flags > after this. I have used this with many SuperMicro servers that use COM2 > and > COM3 because I wanted the entire path (boot loader and kernel) to work, not > the kernel only. Having only the kernel means I can't break into the > loader > prompt to boot a different kernel, single user, etc. > > To clarify, are you using _different_ serial ports for the loader vs the > kernel? That is the use case I considered to be rare. Every single server > I have ever worked with (though not iLO, mostly Dell and SuperMicro) uses > the > same COM port for serial redirection rather for SOL or via actual cables. > I've yet to use a system that, for example, used COM1 for the loader and > COM2 > for the kernel. You are saying that every HP server uses COM1 for the > loader > and COM2 for the kernel? > > -- > John Baldwin > Hello, For Supermicro: The box was in a rack when I started this, so I don't really know how the bios is setup, but as far as I can gather, getting "BTX loader yadda yadda" seems to always happen, on any COM-port. The actual output from loading modules and so on only happens with specific settings. Here ( as in an answer to Dmitry I accidentally sent in private ) things seems to be on different COMs: Setting: comconsole_port=0x2F8 (com2) show loader and goes dark on handoff to kernel comconsole_port=0x3E8 (com3) show no loader and stays dark Neither of them gives me console even after successful boot. The only way to get serial is setting hw.uart.console="br:9600" explicitly. For HP I asked a colleague, and we couldn't exactly remember ( it was quite a while since these boxes were deployed ), but our recollection was that serial setup was left as-is, which seems to be that "vsp" from iLO gives you a tty on 0x2F8. Setting that in comconsole_port in loader.conf works. Both In an effort of standardising stuff, we put hw-dependent stuff in loader.conf, ie hint.uart redirection to have the same entry in the ttys file. Having serial over ipmi/ilo go dark during remote upgrades is scary at best. Seeing that this change has such a potential, how come it is not mentioned in the release notes? Best regards Andreas
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAPS9%2BSuP%2Bi6W5VHkx7gaec=QqKimEwGd1t30sdz8zN6WVcCGiQ>