Page 1 of 1
Kernel too old
Posted: Sat Jul 23, 2011 1:22 am
by DiverGuy
I have several FreeBSD servers that are all cycling though WU's and getting the following errors.
This looping continues for 5 WU's and then shuts down for 1 day.
If I delete all files and start over, (in an effort to force a different WU), it generally happens again.. until eventually a new WU is assigned that does not require the a4 core:
Code: Select all
[01:10:09] Initial: 1B81; + 2191360 bytes downloaded
[01:10:09] Initial: 3189; + 2201600 bytes downloaded
[01:10:09] Initial: 2E98; + 2201948 bytes downloaded
[01:10:09] Verifying core Core_a4.fah...
[01:10:09] Signature is VALID
[01:10:09]
[01:10:09] Trying to unzip core FahCore_a4.exe
[01:10:10] Decompressed FahCore_a4.exe (5803876 bytes) successfully
[01:10:11] + Core successfully engaged
[01:10:16]
[01:10:16] + Processing work unit
[01:10:16] Core required: FahCore_a4.exe
[01:10:16] Core found.
[01:10:16] Working on Unit 01 [July 23 01:10:16]
[01:10:16] + Working ...
[01:10:16] - Calling './FahCore_a4.exe -dir work/ -suffix 01 -checkpoint 15 -verbose -lifeline 78610 -version 602'
FATAL: kernel too old
[01:10:16] CoreStatus = 0 (0)
[01:10:16] Client-core communications error: ERROR 0x0
[01:10:16] Deleting current work unit & continuing...
FATAL: kernel too old
[01:10:16] - Warning: Could not delete all work unit files (1): Core file absent
[01:10:16] Trying to send all finished work units
[01:10:16] + No unsent completed units remaining.
[01:10:16] - Preparing to get new work unit...
[01:10:16] + Attempting to get work packet
[01:10:16] - Will indicate memory of 1023 MB
[01:10:16] - Connecting to assignment server
[01:10:16] Connecting to http: // assign. stanford. edu:8080/
[01:10:17] Posted data.
[01:10:17] Initial: 4A81; - Successful: assigned to (129.74.85.15).
[01:10:17] + News From Folding@Home: Welcome to Folding@Home
[01:10:17] Loaded queue successfully.
[01:10:17] Connecting to http: // 129.74.85.15 :8080/
Re: Kernel too old
Posted: Sat Jul 23, 2011 10:15 pm
by Hyperlife
What version of FreeBSD are you running, and which linux_base port do you have installed?
Re: Kernel too old
Posted: Sat Jul 23, 2011 11:13 pm
by DiverGuy
Hyperlife wrote:What version of FreeBSD are you running, and which linux_base port do you have installed?
FreeBSD 7.3-RELEASE-p4
linux_base-fc-4_16
FreeBSD 7.3-RELEASE-p6
linux_base-fc-4_16
those are two that i know are having problems.
Note: they run fine with cores other than a4
59036 folding 1 138 20 26064K 19172K CPU1 1 22.0H 98.19% FahCore_78.exe
63664 folding 1 137 20 26064K 19164K RUN 0 22.0H 96.00% FahCore_78.exe
***
And, FreeBSD 8.1-RELEASE with the same linux_base-fc-4_16
seems to work just fine
76315 folding 1 138 20 216M 20808K RUN 0 47.0H 100.00% FahCore_a4.exe
72033 folding 1 138 20 217M 21168K CPU3 3 47.0H 100.00% FahCore_a4.exe
67712 folding 1 138 20 202M 10052K CPU2 2 46.8H 100.00% FahCore_a4.exe
64019 folding 1 138 20 202M 9824K CPU1 1 46.8H 100.00% FahCore_a4.exe
Re: Kernel too old
Posted: Sat Jul 23, 2011 11:44 pm
by bruce
I'm not a Linux expert, but I do have a guess. This might be a problem with brandelf.
At one time there was (and there may still be) a command-line parameter -freeBSD which caused the client to properly brand any FahCore that it downloaded to be the proper ELF type. You probably can fix it manually, too.
Re: Kernel too old
Posted: Sun Jul 24, 2011 12:13 am
by Hyperlife
My guess is that Core A4 requires a minimum 2.6 Linux kernel, and 7.3/linux_base-fc4 only supports 2.4.2.
The default for FreeBSD 8.0+ is now linux_base-f10, which emulates a 2.6.16 kernel.
You may need to upgrade your 7.3 systems to at least 8.0 (preferably 8.2) and install linux_base-f10 to run A4 WUs. Your 8.1 rig is probably reporting the 2.6.16 kernel version even though you have the fc4 Linux port installed -- you should be running the f10 port with 8.1 anyway.
Another possibility is to simply tell the 7.3 system that it supports 2.6.16, though I'm not sure if this will actually work. As root, try the following on a 7.3 system:
Code: Select all
# sysctl compat.linux.osrelease=2.6.16
brandelf would be a solution if the client couldn't recognize the A4 core's filetype. Obviously, the client can recognize the filetype, since it's reporting an issue with the kernel version.
Incidentally, if you want to run SMP on a FreeBSD system, running Wine with the Windows client works fine. I've replaced all my uniprocessor clients on my FreeBSD rigs with Wine-emulated SMP clients.
Re: Kernel too old
Posted: Sun Jul 24, 2011 12:51 am
by DiverGuy
Yes, I was (and am) unable to install the linux_base-f10 emulator on FreeBSD 7.3
and the a4 core probably requires newer than 2.4 linux kernel
I'll give the sysctl mod a shot.. cant do much more than melt it all down!
**
But if it doesn't work.. and even if it does.. perhaps there should be an option to not accept core-a4 WU's so that we dont just shut down systems due to this issue.
Re: Kernel too old
Posted: Sun Jul 24, 2011 12:57 am
by DiverGuy
the sysctl mod allowed me to install the linux_base-f10
I returned to 'screen' and restarted the folding jobs that were active (currently both are core_78's)
They should both be completed soon, so we'll see if the A4's run now.
Re: Kernel too old
Posted: Sun Jul 24, 2011 1:08 am
by DiverGuy
Switching to the linux_base-f10 seems to be working on another 7.3 box that downloaded the a4 core.
[01:07:42] Working on Unit 02 [July 24 01:07:42]
[01:07:42] + Working ...
[01:07:42] - Calling './FahCore_a4.exe -dir work/ -suffix 02 -checkpoint 15 -verbose -lifeline 13376 -version 602'
[01:07:42]
[01:07:42] *------------------------------*
[01:07:42] Folding@Home Gromacs GB Core
[01:07:42] Version 2.27 (Dec. 15, 2010)
[01:07:42]
[01:07:42] Preparing to commence simulation
[01:07:42] - Looking at optimizations...
[01:07:42] - Created dyn
[01:07:42] - Files status OK
[01:07:42] - Expanded 53028 -> 204940 (decompressed 386.4 percent)
[01:07:42] Called DecompressByteArray: compressed_data_size=53028 data_size=204940, decompressed_data_size=204940 diff=0
[01:07:42] - Digital signature verified
[01:07:42]
[01:07:42] Project: 7005 (Run 3, Clone 35, Gen 10)
[01:07:42]
[01:07:42] Assembly optimizations on if available.
[01:07:42] Entering M.D.
G R O M A C S (-:
Groningen Machine for Chemical Simulation
VERSION 4.5.3 (-:
Thanks.
Re: Kernel too old
Posted: Sun Jul 24, 2011 2:12 am
by Hyperlife
Glad it works. You should add compat.linux.osrelease=2.6.16 to /etc/sysctl.conf on the 7.3 systems so that the version change setting survives a reboot.
Re: Kernel too old
Posted: Sun Jul 24, 2011 3:19 pm
by DiverGuy
ofc.
thank you.