View previous topic :: View next topic |
Author |
Message |
hbp4c n00b
Joined: 17 Apr 2002 Posts: 46 Location: Charlottesville, Va
|
Posted: Thu May 12, 2005 10:08 pm Post subject: |
|
|
I posted in February that I was having problems with my cobalt raq 2 freezing up whenever large amounts of data was being transferred to or from it. I finally tracked down the problem today (the machine was put aside for a while and I thought about other problems instead ).
If you use hdparm on a raq, I would advise against using the "using_dma" flag (-d1) which hdparm enables by default when you start the hdparm service in /etc/init.d/.
Here is the /etc/conf.d/hdparm that I use and seems to work ok:
Code: | all_args="-c3 -d0 -k1 -m1" |
Hope this helps someone... |
|
Back to top |
|
|
DingbatCA Guru
Joined: 07 Jul 2004 Posts: 384 Location: Portland Or
|
Posted: Wed Jun 15, 2005 5:06 pm Post subject: |
|
|
Just a little bug on my cobalt raq2. This does not effect my system as best as i can tell; would just like to have a nice clean boot.
any ideas?
* Setting user font...Couldnt open /dev/tty1
Couldnt open /dev/tty2
Couldnt open /dev/tty3
Couldnt open /dev/tty4
Couldnt open /dev/tty5
Couldnt open /dev/tty6
Couldnt open /dev/tty7
Couldnt open /dev/tty8
Couldnt open /dev/tty9
Couldnt open /dev/tty10
Couldnt open /dev/tty11
* Failed to set user font
[ !! ]
* Loading key mappings...Couldnt get a file descriptor referring to the console
* Error loading key mappings
[ !! ]
* Bringing lo up... [ ok ] |
|
Back to top |
|
|
hbp4c n00b
Joined: 17 Apr 2002 Posts: 46 Location: Charlottesville, Va
|
Posted: Wed Jun 15, 2005 6:01 pm Post subject: |
|
|
DingbatCA,
Edit /etc/inittab and remove or comment out the lines that start mingetty. This will disable the warnings.
Code: | #1:2345:respawn:/sbin/mingetty tty1
#2:2345:respawn:/sbin/mingetty tty2
#3:2345:respawn:/sbin/mingetty tty3
#4:2345:respawn:/sbin/mingetty tty4
#5:2345:respawn:/sbin/mingetty tty5
#6:2345:respawn:/sbin/mingetty tty6 |
|
|
Back to top |
|
|
DingbatCA Guru
Joined: 07 Jul 2004 Posts: 384 Location: Portland Or
|
Posted: Thu Jun 16, 2005 12:48 am Post subject: |
|
|
-hbp4c
Hummm, Edited /etc/inittab and remove or comment out the lines that starting agetty, so c1~c6. Still no luck on the start up error. But thats just a little pain, here is the big pain.
When I put the system under a heavy load it hangs.
I saw your post about editing the hdparm and using "all_args="-c3 -d0 -k1 -m1"" Well, modified my hdparm, gave the system a reboot and turned on samba. Started the load test... Hang.
Here is my load test
On the cobalt set up samba and make an open share with at lease 1GB free space.
On my windows system install sandra (sisoftware.net)
make a network drive to the open share.
run the file system bench make against my mapped network drive. The cobalt then hangs very quickly (1~2min).
Any ideas? Do I need any thing special to use hdparm or is it a Gentoo default
-DingbatCA |
|
Back to top |
|
|
hbp4c n00b
Joined: 17 Apr 2002 Posts: 46 Location: Charlottesville, Va
|
Posted: Thu Jun 16, 2005 2:47 am Post subject: |
|
|
For the first problem, you may also check to be sure that /etc/init.d/consolefont isn't being loaded on boot up. It is enabled by default when you install baselayout... so an rc-update del consolefont should fix it.
Sorry I mentioned mingetty, i was looking at a fedora machine since I deleted the lines in inittab on my cobalt.
As for the hangs...
You may try to disable hdparm completely via an rc-update del hdparm and see if the machine still has problems with freezing. Don't forget, after you remove hdparm, you should reboot to make sure that nothing is touched on the disk.
The "all_args="-c3 -d0 -k1 -m1"" settings work for me, but I am using a western digital 40GB disk that obviously did not come with the cobalt originally. If you're still using the OEM drive that came with the cobalt, one of the other settings may be giving you troubles.
Good luck. |
|
Back to top |
|
|
hbp4c n00b
Joined: 17 Apr 2002 Posts: 46 Location: Charlottesville, Va
|
Posted: Thu Jun 16, 2005 3:05 am Post subject: |
|
|
Since the thread is bumping along a little again, i'll ask:
Has anyone tried a recent emerge -ua world directly on their cobalt? It seems like almost every package I try to compile now has problems. FYI, I'm using the stage 3 install with no updated software (emerge bombs while updating ncurses which is first on the list. I've also tried to manually emerrge a later {stable} version of gcc, automake and even bison, but all fail to compile). However, I know that gcc is working at least partially since samba and apache compiled ok.
Is everyone else cross compiling on another machine?
Any thoughts?
Howard |
|
Back to top |
|
|
DingbatCA Guru
Joined: 07 Jul 2004 Posts: 384 Location: Portland Or
|
Posted: Thu Jun 16, 2005 4:02 am Post subject: |
|
|
Cool, fixed the font issue, thx for the fix.
Still one left at boot.
* Error loading key mappings [ !! ]
I think i have a setting wrong in the rc.conf
But on the other hand, fixed the crashing issue.
"rc-update add hdparm boot" Thx hbp4c
Now tuning my settings last test took 45 min and came in at a finished speed of 2MB/sec speed... I no that my old 8.4GB hard disk is slow, but I am sure It can do better then that.
Do you no of any good links for what all the settings are in "hdparm"
Hummm, is there a max HDD (Hard Disk Drive) size on these cobalts? 8GB, 136GB... I no it can run a 120GB. Has any one tried any thing better then 136 on the IDE bus? |
|
Back to top |
|
|
hbp4c n00b
Joined: 17 Apr 2002 Posts: 46 Location: Charlottesville, Va
|
Posted: Thu Jun 16, 2005 1:14 pm Post subject: |
|
|
Here's the results from the built-in hdparrm tests to compare:
Code: | hadhafang root # hdparm -Tt /dev/hda
/dev/hda:
Timing buffer-cache reads: 72 MB in 2.06 seconds = 34.88 MB/sec
Timing buffered disk reads: 10 MB in 3.54 seconds = 2.82 MB/sec
hadhafang root #
|
I am using the string listed above for my hdparm settings.
This thing is certainly not a speed demon.
h |
|
Back to top |
|
|
DingbatCA Guru
Joined: 07 Jul 2004 Posts: 384 Location: Portland Or
|
Posted: Thu Jun 16, 2005 3:08 pm Post subject: |
|
|
Humm, so i should be able to to better then 2MB
cobalt root # hdparm -Tt /dev/hda
/dev/hda:
Timing cached reads: 80 MB in 2.04 seconds = 39.23 MB/sec
Timing buffered disk reads: 24 MB in 3.27 seconds = 7.35 MB/sec
Just for fun i hit my home server :-)
whitequeen ~ # hdparm -Tt /dev/md0
/dev/md0:
Timing cached reads: 672 MB in 2.00 seconds = 335.55 MB/sec
Timing buffered disk reads: 166 MB in 3.01 seconds = 55.16 MB/sec |
|
Back to top |
|
|
Redhatter Retired Dev
Joined: 20 Sep 2003 Posts: 548 Location: Brisbane, QLD, Australia
|
Posted: Thu Jun 16, 2005 11:08 pm Post subject: |
|
|
Well.... the keyboard mapping errors are most likely due to the lack of a local console & keyboard. If I'm not mistaken... there's no socket anywhere on a Cobalt server where you could plug a keyboard or monitor. So you may as well take keymaps out of the boot sequence. (try rc-update -d keymaps)
As for the speed issue... I've got a 10GB IBM Deskstar (IDE) and a couple of SCSI drives plugged into my Qube (via an Adaptec AHA2940AU in the PCI slot)...
Code: | qube root # dmesg | grep ^hda:
hda: IBM-DPTA-371020, ATA DISK drive
hda: max request size: 128KiB
hda: 20028960 sectors (10254 MB) w/1961KiB Cache, CHS=19870/16/63, UDMA(33)
qube root # cat /proc/scsi/scsi
Attached devices:
Host: scsi0 Channel: 00 Id: 00 Lun: 00
Vendor: SEAGATE Model: ST150176LW Rev: 0002
Type: Direct-Access ANSI SCSI revision: 02
Host: scsi0 Channel: 00 Id: 02 Lun: 00
Vendor: SEAGATE Model: ST150176LW Rev: 0002
Type: Direct-Access ANSI SCSI revision: 02
Host: scsi0 Channel: 00 Id: 04 Lun: 00
Vendor: IBM Model: DDRS-39130W Rev: S92A
Type: Direct-Access ANSI SCSI revision: 02
qube root # for dev in hda sda sdb sdc; do hdparm -tT /dev/$dev; done
/dev/hda:
Timing cached reads: 80 MB in 2.04 seconds = 39.27 MB/sec
Timing buffered disk reads: 34 MB in 3.18 seconds = 10.70 MB/sec
/dev/sda:
Timing cached reads: 80 MB in 2.04 seconds = 39.29 MB/sec
Timing buffered disk reads: 22 MB in 3.02 seconds = 7.29 MB/sec
/dev/sdb:
Timing cached reads: 80 MB in 2.04 seconds = 39.22 MB/sec
Timing buffered disk reads: 22 MB in 3.01 seconds = 7.32 MB/sec
/dev/sdc:
Timing cached reads: 80 MB in 2.04 seconds = 39.23 MB/sec
Timing buffered disk reads: 24 MB in 3.20 seconds = 7.50 MB/sec |
Note I have DMA enabled. I don't know what might be causing the instability you've been seeing... but I haven't had any problems here with DMA enabled. _________________ Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...
Gentoo/MIPS Cobalt developer, Mozilla herd member. |
|
Back to top |
|
|
DingbatCA Guru
Joined: 07 Jul 2004 Posts: 384 Location: Portland Or
|
Posted: Fri Jun 17, 2005 9:19 pm Post subject: |
|
|
With out a question It is DMA. So you think you cube is so stable :-)
Here is a great test I use against allot of Linux systems.
Get samba installed on the target (Cobalt RaQ2 in my case) Use a very striped down smb.conf
Install Sandra on a windows box that can map a network drive to your target. (sisoftware.net)
Map a network drive to your target and run Sandras file system bench mark against the mapped network drive.
I have not seen a cobalt RaQ2 that is able to run this with DMA enabled on an IDE HDD. |
|
Back to top |
|
|
DingbatCA Guru
Joined: 07 Jul 2004 Posts: 384 Location: Portland Or
|
Posted: Sat Jun 18, 2005 2:40 am Post subject: |
|
|
MOHAHAHA
Victory is MINE!!
OK, the DMA issue, I have it solved. Now running with DMA on and no more crashing issues. Normally when I see things like that on an X86 box, it takes a bios update to fix. So I moved from colo 1.13 to colo 1.19. Ran my test. The system hung. At this point I was guessing it was a bios problem. I did not think Sun Microsystems would release a box like this with that big an issue.
Re-flashed the old stock Cobalt RaQ2 bios.
Rebooted. Wow I am bright
.
Net booted, placed the colo-chain.elf in /boot as vmlinux.gz.
Rebooted. Wow I am bright
.
Backed up my data in /boot and reformatted it in ext2 rev0. Restored my data.
Rebooted
System came up but did not load my menu.colo. How do I get it to default to reading my menu.colo? Loaded the kernel by hand.
On to my stress test. Nailed the system, load averages over 3. Stress test finished at 2.7MB; my old set of tests finished at 2MB, with out DMA. What a difference. :-) I hope with a faster had drive I can make that move a bit faster. |
|
Back to top |
|
|
Redhatter Retired Dev
Joined: 20 Sep 2003 Posts: 548 Location: Brisbane, QLD, Australia
|
Posted: Sat Jun 18, 2005 8:44 am Post subject: |
|
|
Rename menu.colo to default.colo and things should work. CoLo automatically tries to load that script. _________________ Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...
Gentoo/MIPS Cobalt developer, Mozilla herd member. |
|
Back to top |
|
|
DingbatCA Guru
Joined: 07 Jul 2004 Posts: 384 Location: Portland Or
|
Posted: Mon Jun 20, 2005 6:40 pm Post subject: 64-bit? |
|
|
How is 64-bit support looking on the mips systems (Cobalt RaQ2). Is it stable in any configuration??
If so, what kernel/patches....
DingbatCA |
|
Back to top |
|
|
Redhatter Retired Dev
Joined: 20 Sep 2003 Posts: 548 Location: Brisbane, QLD, Australia
|
Posted: Tue Jun 21, 2005 2:18 am Post subject: |
|
|
DingbatCA:
MIPS systems in general... no problems. All SGI systems (other than the IP22 family) currently supported run MIPS64. IP22's can run MIPS64 kernels too, but it's a bit on the flakey side right just now.
Cobalt Servers... according to this site, support is incomplete. In other words, it may work, but only just... don't expect everything to work, or the ride to be smooth. _________________ Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...
Gentoo/MIPS Cobalt developer, Mozilla herd member. |
|
Back to top |
|
|
DingbatCA Guru
Joined: 07 Jul 2004 Posts: 384 Location: Portland Or
|
Posted: Tue Jun 28, 2005 3:56 pm Post subject: |
|
|
Well, gave 64-bit a test... (no patch's, stock kernel)
System booted just fine, but found that the network card no longer worked. Recompiled and turned the network card into a module. Still no luck.Getting lots of errors in dmesg and the logs. Back to 32 bit mode, every thing works just fine. Well it was fun to try.
DingbatCA |
|
Back to top |
|
|
DingbatCA Guru
Joined: 07 Jul 2004 Posts: 384 Location: Portland Or
|
Posted: Thu Jun 30, 2005 6:11 am Post subject: More speed |
|
|
In my never ending quest to squeeze more speed out of my Cobalt RaQ2; I am now working on the network. The best numbers I can get are about 3.7MB/s. My tests are a bench mark from you windows box to my cobalt. My target is on my cobalt. It's a 64MB ram drive. So I know thats very fast. The only part the seems to be slow is the NIC on the Cobalt RaQ2.
Build up a simple ramdrive on tmp.
mount -t tmpfs tmpfs -o size=64M,nr_inodes=1M /tmp
Point smb.conf file to the ram drive.
Run test. 3.7MB/s
My home server comes in at 9MB/s on a 10/100 connection and testing hard drives, not a ram drive. Have I found the best that the hardware can do?
Any ideas?
DingbatCA |
|
Back to top |
|
|
tag n00b
Joined: 06 Mar 2003 Posts: 55
|
Posted: Wed Jul 13, 2005 3:14 pm Post subject: unit boot failure |
|
|
I have just acquired a raq2 and setup the nfs boot / dhcp to boot the raq from the /nfsroot, but I get:
"UNIT BOOT FAILURE"
this is what I have done:
created /nfsroot
downloaded colo-1.16.tar.gz
gzip -9vc colo-chain.elf > /nfsroot/vmlinux_raq-2800.gz
cd /nfsroot
ln -s . /boot
added the following the DHCP server:
subnet 192.168.20.0 netmask 255.255.255.224 {
# ... usual stuff here ...
# Configuration for a Cobalt Server
# Set the hostname here:
host qube {
# Path to the nfsroot directory.
# This is mainly for when using the TFTP boot option on CoLo
option root-path "/nfsroot";
# Cobalt server's ethernet MAC address
hardware ethernet 00:10:e0:00:a0:e3;
# Server to download image from
next-server 192.168.20.1;
# IP address of cobalt server
fixed-address 192.168.20.2;
# Image to download
# Again, this has more to do with CoLo
filename "vmlinux_raq-2800.gz";
}
}
attached the cable to serial on raq2 - and fired up kermit.
powered on the raq while holding the left and right keys got the boot menu and selected BOTH "boot shell" and "nfs boot".
On the Boot shell - it does nothing but sit there - and on kermit I get :
H·oÿ÷7õ:kÐN78V{ÿ¶fµ:kÙÎvÈæ6³ðfeþ771
the dhcp logs show:
Jul 13 17:09:52 tag dhcpd: BOOTREQUEST from 00:10:e0:00:a0:e3 via eth0
Jul 13 17:09:52 tag dhcpd: BOOTREPLY for 192.168.20.2 to qube (00:10:e0:00:a0:e3) via eth0
Jul 13 17:09:52 tag rpc.mountd: authenticated mount request from raq:600 for /nfsroot (/nfsroot)
Jul 13 17:10:02 tag dhcpd: DHCPDISCOVER from 00:10:e0:00:a0:e3 via eth0
Jul 13 17:10:02 tag dhcpd: DHCPOFFER on 192.168.20.2 to 00:10:e0:00:a0:e3 via eth0
Jul 13 17:10:02 tag dhcpd: DHCPREQUEST for 192.168.20.2 (192.168.10.101) from 00:10:e0:00:a0:e3 via eth0
Jul 13 17:10:02 tag dhcpd: DHCPACK on 192.168.20.2 to 00:10:e0:00:a0:e3 via eth0
and the same if I choose nfs boot - but this time I get "UNIT BOOT FAILURE"
what am I doing wrong ?
Thanks |
|
Back to top |
|
|
Redhatter Retired Dev
Joined: 20 Sep 2003 Posts: 548 Location: Brisbane, QLD, Australia
|
Posted: Thu Jul 14, 2005 8:43 am Post subject: |
|
|
Looks like kermit isn't configured correctly. You need to tell your serial program to use 115200 baud, 8-bit, no parity bits, 1 stop bit. Either that, or your null modem cable has issues.
Push comes to shove, try minicom... I use that program personally, and know it works. screen also works well with these machines. (but IMHO, minicom is easier to use )
The other alternative you can try... is to use CoLo 1.19. In the tarball, there's a little utility called ncon, it's a network console client. CoLo 1.19 BTW is also in portage (under ~mips).
On the server:
1. Compile ncon by changing to the tools/ncon directory and running make.
2. Start ncon, giving it the IP address of your Raq2
3. Proceed with creating the vmlinux.gz_raq-2800.gz as per the guide
... Then... on the Raq2...
1. Boot the machine by holding the left/right arrow buttons
2. Select network console from the menu.
You should then be able to interact with the shell via the DHCP server running ncon.
When you get to the execute line... instead, type execute dhcp telnet. That should start up a telnet server from which you can install Gentoo.
This is all due to go into the guide, but I've still got to work out how to do netconsole with the kernel...which should allow Qube 2700 users to install Gentoo easily. _________________ Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...
Gentoo/MIPS Cobalt developer, Mozilla herd member. |
|
Back to top |
|
|
tag n00b
Joined: 06 Mar 2003 Posts: 55
|
Posted: Thu Jul 14, 2005 8:52 am Post subject: ncon is cool |
|
|
Thanks for the info - ncon works well ... I got it to boot from the colo1.19 version but, I was trying to install the kernel by using the :
nfs 192.168.20.1 /nfsroot vmlinuz.gz
execute
I will try the :
execute dhcp telnet
option
the kermit was configured correcly - it is just the cable was a bit of an issue
Thanks. |
|
Back to top |
|
|
tag n00b
Joined: 06 Mar 2003 Posts: 55
|
Posted: Fri Jul 15, 2005 6:00 am Post subject: |
|
|
hmm - maybe I am not doing this correctly - but when I type :
> execute dhcp telnet
I get:
no image loaded
this is from the ncon command line ...
any ideas ?
Thanks |
|
Back to top |
|
|
Redhatter Retired Dev
Joined: 20 Sep 2003 Posts: 548 Location: Brisbane, QLD, Australia
|
Posted: Sun Jul 17, 2005 7:56 am Post subject: |
|
|
Did it load the kernel correctly? You still need to load the kernel (using either the nfs or tftp commands). _________________ Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...
Gentoo/MIPS Cobalt developer, Mozilla herd member. |
|
Back to top |
|
|
tag n00b
Joined: 06 Mar 2003 Posts: 55
|
Posted: Sun Jul 17, 2005 12:38 pm Post subject: |
|
|
hmm - I thought that it already done that, when it boots into the network console ?
What is the name of the kernel ?
Where can I get this kernel ?
Methinks I have missed a command here ?
Thanks again -
Tonino |
|
Back to top |
|
|
Redhatter Retired Dev
Joined: 20 Sep 2003 Posts: 548 Location: Brisbane, QLD, Australia
|
Posted: Sun Jul 17, 2005 1:34 pm Post subject: |
|
|
You'll find all the cobalt-related files at http://dev.gentoo.org/~redhatter/mips/cobalt/. I'll put up notices if/when files get moved... in general, I don't keep files if they've been pushed out to mirrors.
http://www.gentoo.org/doc/en/handbook/handbook-mips.xml?part=1&chap=2#doc_chap5 explains the netboot process step by step.
As I mentioned above though, to get the netboot to start a telnet daemon automatically, type execute dhcp telnet instead. This is an, as yet, undocumented feature, as it still needs more testing.
Basically... earlier, where you typed:
Code: | nfs 192.168.20.1 /nfsroot vmlinuz.gz
execute |
Instead... type:
Code: | nfs 192.168.20.1 /nfsroot vmlinuz.gz
execute dhcp telnet |
_________________ Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...
Gentoo/MIPS Cobalt developer, Mozilla herd member. |
|
Back to top |
|
|
tag n00b
Joined: 06 Mar 2003 Posts: 55
|
Posted: Mon Jul 18, 2005 6:58 am Post subject: |
|
|
ahh - I did it the wrong way ...
I did :
nfs 192.168.20.1 /nfsroot vmlinuz.gz
execute
then :
execute dhcp telnet
thanks for the clarification |
|
Back to top |
|
|
|