Archive for August, 2005

Split views in DNS

Friday, August 26th, 2005

By running a DNS query against the outside interface of my firewall (65.40.217.234), I can see that my DNS configuration is not correct.  Specifically, named thinks queries originating at the firewall (192.168.1.1) are for the internal view and returns a worthless address:

[root@triumph etc]# dig @65.40.217.234 www.hapgoods.com

…

;; ANSWER SECTION:

www.hapgoods.com.

259200  IN

CNAME   ducati.hapgoods.com.

ducati.hapgoods.com.

259200  IN

A            192.168.1.20

…

(BTW, here is a useful command to determine the IP address from which the rest of the world sees you originating:

[root@ducati named]# curl -s http://checkip.dyndns.org/ | grep -i address

)

So, this raises some interesting questions.  First, from which address would a true external query appear to originate?  Does the named determination of the origin address examine the IP header?  To gain some further insight, I looked into the debug output from named as follows:

[root@ducati log]# touch /var/named/chroot/var/named/named.run

[root@ducati log]# chown named:named /var/named/chroot/var/named/named.run

[root@ducati log]# rndc querylog

[root@ducati log]# rndc trace

[root@ducati log]# rndc trace

[root@ducati log]# tail -f /var/named/chroot/var/named/named.run

This allowed me to clearly see that named considered 192.168.1.1 as belonging to the private view -despite having explicitly configured this address in the public view:

Screen clipping taken: 26/08/2005, 16:43

I couldn’t find anything pertinent on the Internet.  Then it occurred to me that perhaps the sequence of statements was more important than the explicitness of the entries in /etc/named.conf.  Background: some configuration files automatically increase the priority of a configuration statement as the explicitness of the IP address increases.  So host addresses are automatically more important than network addresses.  But this is not the case with named.  So I simply switched the order of the public and private views in the configuration file and suddenly my “external” queries began getting the data from the public view.

ACPI and other hardware management technologies on the Proliant 3000

Thursday, August 25th, 2005

I did disable ACPI on suzuki by adding the following parameter to end of the first kernel entry in /boot/grub/grub.conf:

acpi=off

Now, when suzuki boots, I no longer see the error message noted above.  However, further on in the boot messages, I do see these errors/warnings:

…

DMI not present.

Using APIC driver default

…

…DMI (Desktop Management Interface) not appropriate for server

ENABLING IO-APIC IRQs

..TIMER: vector=0×31 pin1=2 pin2=0

..MP-BIOS bug: 8254 timer not connected to IO-APIC

…trying to set up timer (IRQ0) through the 8259A …

….. (found pin 0) …works.

…

This problem is like a rash on the dmesg output postings on the Internet.  I don’t think I am alone, and I conclude that it is not a serious problem.  I also don’t think there is a way to bypass the ERROR stage of discovering the APIC configuration.

ACPI: Subsystem revision 20050309

ACPI: Interpreter disabled.

…

This website has some glimmer of hope: http://64.233.167.104/search?q=cache:8MXU7mYpaFkJ:gentoo-wiki.com/HARDWARE_Compaq_Proliant_6500+grub+mbr+/boot+partition+fdisk+compaq+f10&hl=en

Linux Plug and Play Support v0.97 (c) Adam Belay

pnp: PnP ACPI: disabled

usbcore: registered new driver usbfs

usbcore: registered new driver hub

…

…Perhaps I should disable USB support (nousb) and ISA PnP support (noisapnp)

PCI BIOS passed nonexistent PCI bus 1!

PCI: Device 0000:00:00.0 not found by BIOS

PCI: Device 0000:00:0f.0 not found by BIOS

PCI: Device 0000:00:11.0 not found by BIOS

…

…This “problem” is normal with ProLiants.  It really means “No driver found for these PCI devices” which are the ASICs for System Management.  See http://www.van-dijk.net/mailarchive/vandijk0108/0010.html

apm: BIOS not found.

…

…Perhaps I should disable APM support (noapm, or apm=off?  I think it is apm=off)

Linux agpgart interface v0.101 (c) Dave Jones

agpgart: Detected a Serverworks chipset but could not find the secondary device.

agpgart: Detected a Serverworks chipset but could not find the secondary device.

…

…No AGP in Compaq ProLiant 3000s -can I disable AGP probing?  This link (http://www.kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.5-rc3) shows that I am not the only to report this problem.  Maybe I can disable the loading of the agp module -but so far, short of renaming the module itself (/lib/modules/`uname -r`/…/….ko), I don’t see how.  Same goes for bluetooth.

See http://www.atconsultancy.nl/atmission/faq.html

http://www.ibiblio.org/peanut/Kernel-2.6.12.2/kernel-parameters.txt

Summary: I would REALLY like to know definitively which technologies are supported on the ProLiant 3000 various iterations.

Technology

6/333

6/550

6/600

DMI

No

No

No

APM

No

No

No

USB

No

No

No

ACPI

No

No

Yes*

AGP

No

No

No**‚

Compaq Advanced System Management (ASM) Controller

No

Yes

Yes

*but with invalid RSDP values

‚**but detected as such

Regarding ACPI, The only guidance I have really found is again Richard Black’s site (http://www.cpqlinux.com/acpi-howto.html#proliant).  But even it is not clear on the specifics of the ProLiant 3000.

ACPI Support on Proliant 3000

Thursday, August 25th, 2005

While inspecting the output of the dmesg command, I note that suzuki is generating kernel startup errors:

dmesg

ACPI: Unable to locate RSDP

Some interesting information on ACPI (Advanced Configuration and Power Interface) can be found here:

http://portal.suse.com/sdb/en/2002/10/81_acpi.html

http://acpi.sourceforge.net/index.html

http://www.tldp.org/HOWTO/ACPI-HOWTO/aboutacpi.html

I find it interesting that ACPI was introduced by Compaq in 1996 (around the time of the ProLiant 3000) yet it does not seem to be supported (http://h200001.www2.hp.com/bc/docs/support/SupportManual/c00208465/c00208465.pdf). In which case, I should disable the probing for ACPI with the kernel parameter acpi=off (/boot/grub/grub.conf). I wonder if the newer ProLiant 3000s (triumph and bmw) will also lack ACPI support?

UPDATE (26/08/2005): triumph (a 6/600) does have ACPI -but with caveats. See later posting on this subject…

Differences between versions of the Proliant 3000

Wednesday, August 17th, 2005

Today I installed the RILOE boards in suzuki and triumph.  While doing so, I noted the following (mainly mechanical) differences between suzuki (a 6/333) and triumph (a 6/600):

  1. Slot CD versus tray CD
  2. Slot screws versus Torx screws in front panel
  3. Small metal indentations versus large plastic indentation on side panel
  4. Interlock switch versus no interlock switch

How to transition from Domain Direct to my network

Wednesday, August 10th, 2005

First there was the plan, and the plan was good:

  1. Open the DNS port on the firewall (currently telnet 65.40.217.234 25 name fails with a “Connection Refused” message.)
  2. Mimic the DNS entries of Domain Direct.  I can research their entries with dig.  I want my internal DNS view to match the existing DD service.
  3. Add my name server to the list of name servers

Proliant 3000 #3 Details

Monday, August 1st, 2005

This past weekend I tore into the new ProLiant. First, it was a bit bent up -but nothing that could not be fixed with a hammer and vise grip pliers. I pseudo-converted it to a tower model from a rack mount model using some of the left-over parts from Matt’s PL3K (suzuki). The noted differences between the two:

Front Bezel

Rack: Wider due to rack ears

Tower: Normal width with intake grill and cut-out for Serial Number.

Serial Number Tag

Rack: On rear, I believe

Tower: On front, in bezel cut-out.

Front Air Intake

Rack: Embedded plastic grill

Tower: Grill part of metal bezel

DIP Switch (Why?)

Rack: Switch 5, Block Two: Off

Tower: Switch 5, Block Two: On

Base (a bitch to mount -requires mostly removing drive cage)

Rack: None

Tower: Wide, holds casters

Casters

Rack: None. Duh.

Tower: Four -fit into Base.

Bottom Cover

Rack: None

Tower: Large Metal panel with door pin holes to fill out rectangular shape.

Door

Rack: None, duh.

Tower: One.

Integrated Management Display

Rack: Oriented 90 degrees to CD ROM drive

Tower: Oriented same as CD ROM drive