The Linux Help and Tutorial Thread

[NKrader]

Historically wireless has been a weakness of Linux, mostly because the hardware producers don't see the point in sharing their info with the Linux driver developers. That leaves them in a position where they have to reverse engineer drivers or write wrapper scripts and use Windows drivers. Ethernet networking has been a strength, with the Linux networking stack historically being so much stronger than the Windows one that Microsoft actually stole it ... and they have the f'ing hide to accuse Linux of violating their patents.

Anyway, enter this command into a terminal.

lspci > lspci-output.txt

This will send the output of the lspci command to a text file in your current working directory (your home directory if you opened a fresh terminal to do it) so you can open it up with a graphical text editor and copy etc more easily. Copy and paste the contents up here. It will list off every device attached to the PCI bus, both internally (like USB controllers) and externally (like your wifi card). That way we can see what devices the OS is seeing.

im running server with no gui

[D_A]

By any chance did you install ssh-server when you installed the system?

If you can get it online via ethernet one of us might be able to remote in and do some of this for you.

[NKrader]

By any chance did you install ssh-server when you installed the system?

If you can get it online via ethernet one of us might be able to remote in and do some of this for you.

nope didnt install ssh-server (thats the only ssh thing I saw options.)

and the only internet i have access too is my phone.. which i could tether, being that its usb it seems like it would work even easier, i will try that tonight.

i need a movieman style 2-3hour tech support phone call too bad he doesnt play the linux game LoL


linux really does interest me, and i really want to start learning it as it is one of few things that I cant really manage to navigate with ease.

[D_A]

One thing I've noticed is that it's easier to do some things in Linux than to explain them, especially when you're trouble shooting. Sometimes I do wonder if penguins are all autistic-spectrum types.

USB tethering is not something I've tried and honestly while they system will recognise the device I don't know if it will use it without specific drivers that may or may not be included by default in your install.

[BlindFreddie]

... the only internet i have access too is my phone.. which i could tether, being that its usb it seems like it would work even easier, i will try that tonight.
linux really does interest me, and i really want to start learning it as it is one of few things that I cant really manage to navigate with ease.

There are lots of sites out there with Linux Q & As. linuxquestions.org is one. Remember too, The Duck is your friend.

Example: http://duckduckgo.com/?q=tether+phone+via+usb+in+linux

---
My WCG linux question: Has anyone here compared crunching speeds for linux vs windoze for the various WCG science programs?
We know that with VINA WUs (FAHV), linux is heaps faster - around +40%. What about the others, ordinary AutoDock (faah) in particular? It's hard to assess those, because their runtimes are so variable.

I'm currently trying running MCM under windoze on some cores, and FAAH in a VM under linux on the remaining cores. The FAHV WUs run really fast, but you also get faah WUs that run slower, and you can't select just 1 type. If you abort the faah WUs in your cache or abort them during download, WCG issues 2 replacement repair WUs for each. You'll get to run only your fast FAHV WUs, but you will have caused lots of wasted crunching of extra faah WUs. WCG could fix it, but I'm sure they won't.

Unless perhaps LOTS of people ask for it to happen ... (??)

[D_A]

I don't think the autodock units are "broken" so they can't be "fixed". The Vina work uses a different computation method that just happens to be extremely efficient under Linux. Could Autodock be optimised better? Possibly. Though it's not as simple as just setting a couple of flags and recompiling. Windows and Linux use different memory structures so the code would need to be optimised to make suit while still producing consistent results.
As for speed, Linux is typically faster on every project, but in some cases it's only barely and the difference is less than the performance increase achieved by switching from a mechanical HDD to a SSD. HCC and FAHV are the stand out examples, the latter being SO much better it pushes Linux machines' points output up significant'y compared to Windows machines on the same project. Most project still award more points to Windows machines, however, due to the disparity between BOINC benchmarks on Windows and Linux. There has been significant discussion about that in the past but I don't think it's ever been resolved.

[BlindFreddie]

I don't think the autodock units are "broken" so they can't be "fixed".

Sorry, D_A, but you misunderstood me. I didn't mean to imply that the AutoDock WUs have a problem, other than that they don't show a huge speed advantage when run under Linux.
I want to run only FAHV WUs under Linux, with other WU types under Windows, to maximise throughput of my machines.
But the Linux machines have to accept AutoDock WUs as well as FAHV ones. That's what I'd like fixed.
WCG could:
1) (easy) Divert nearly all of the faah WUs sent out, to Windows and Mac boxes, sending any needed extra FAHV ones to Linux boxes, or
2) (harder) Run the FAAH project as 2 sub-projects, letting us choose the WU types that we want. The FAAH project would have 2 selection flags on the device profile pages, 1 for AutoDock, the other for VINA.

Enabling the diversion of WUs of various types to the machines on which they run fastest would enable more efficient use of our crunching resources, and raise WCG's total throughput.

Thanks for the info that most of the programs run faster under Linux. I thought that it was the opposite, except for VINA.
ATM, I'm running all machines with Win7-x64 native, and Debian 7.2 in VMs under VMWare. I will allocate more cores to the Linux VMs and fewer to Windows.
I tried VirtualBox, but it seems rather inefficient. OTOH the overhead of WMWare seems to be negligible.
Without me having to dig for the info, does anyone know how to get the VMs to autostart at host system boot time?

[NKrader]

why run virtual? not just run full fledged linux install?

[D_A]

Sorry, D_A, but you misunderstood me. I didn't mean to imply that the AutoDock WUs have a problem, other than that they don't show a huge speed advantage when run under Linux.
I want to run only FAHV WUs under Linux, with other WU types under Windows, to maximise throughput of my machines.
But the Linux machines have to accept AutoDock WUs as well as FAHV ones. That's what I'd like fixed.
WCG could:
1) (easy) Divert nearly all of the faah WUs sent out, to Windows and Mac boxes, sending any needed extra FAHV ones to Linux boxes, or
2) (harder) Run the FAAH project as 2 sub-projects, letting us choose the WU types that we want. The FAAH project would have 2 selection flags on the device profile pages, 1 for AutoDock, the other for VINA.

Enabling the diversion of WUs of various types to the machines on which they run fastest would enable more efficient use of our crunching resources, and raise WCG's total throughput.

Thanks for the info that most of the programs run faster under Linux. I thought that it was the opposite, except for VINA.
ATM, I'm running all machines with Win7-x64 native, and Debian 7.2 in VMs under VMWare. I will allocate more cores to the Linux VMs and fewer to Windows.
I tried VirtualBox, but it seems rather inefficient. OTOH the overhead of WMWare seems to be negligible.
Without me having to dig for the info, does anyone know how to get the VMs to autostart at host system boot time?

The catch is, as far as I know FAVH is the only project that SCORES better on Linux. The BOINC benchmarks return higher on Windows, or certainly used to, and then take more time to do the work units. Since Windows results are compared to Windows results, they tend to return more points. While Linux tends to do the work faster it also benchmarks lower (benchmarks more aggressively optimised by default in Windows compilers than in Linux compilers) so less time with lower benchmark gives less points, even though more work gets done. VINA is the first science app that is SO much faster under Linux that it scores heavily in Linux's favour. HCC was up to 50% faster on Linux as well, but returned less points. Generally speaking, Linux has not been the OS of choice for points hounds until FAVH. It's advantages have been more to do with the lack of Microsoft Tax in setting up large numbers of machines and stability, not needing to be restarted (or doing so automatically) every other Tuesdays due to automatic updates.

I concur with NKRaider. Why not just run native Linux instead of running VMs? It's more efficient than running in a VM and less complicated to administer.

[NKrader]

If I was able to get on ethernet webs and have someone remote in should I reinstall with ssh server?

Looks like the parts for the Linux dedicated will be here on Tuesday so that night or most likely wed I could arrange to use a "real" network, but I will be testing USB tether after it all gets here.

[D_A]

I'd suggest installing the SSH server on install, yes. To get someone to log in you'll need to know your external IP address, possibly open a port in the router firewall (that's assuming the router even has a firewall), and give the person you're letting in the username and password for the system.

[NKrader]

I'd suggest installing the SSH server on install, yes. To get someone to log in you'll need to know your external IP address, possibly open a port in the router firewall (that's assuming the router even has a firewall), and give the person you're letting in the username and password for the system.

Yep, figured out how to tether my phone router so I have ethernet now at home when I need it.
And yes my router has firewall asus n66u

[D_A]

If you decide to do this, you might want to enable port forwarding in your router and forward port 22 to the IP of the machine in question. If it's the only machine on the local network just turning off the firewall temporarily might also work, but that's not the ideal solution.

[PoppaGeek]

How To install and configure command line only Ubuntu server and boinc-client, and manage it from another machine.

This guide assumes you will install Ubuntu Server and manage and control Boinc on the server from a second PC, running Linux with a desktop GUI or running Windows. The LTS (Long Term Support)version is recommended.

To see if your hardware, especially network gear works OK with Ubuntu you can download a Live CD/DVD of Ubuntu, boot from it and see if you have internet access by opening a browser and surfing the web, or if other hardware does/does not work OK.


How to Install Ubuntu Server- A step-by-step picture guide and video on how to install Ubuntu Server. (In step 11 - Check OpenSSH Server)


After install and reboot:

Update system

sudo apt-get update

If you did not check OpenSSH Server during install or not sure if you did:

sudo apt-get install openssh-client

sudo apt-get install openssh-server

Install boinc-client

sudo apt-get install boinc-client

DO NOT install boinc-manager as it will then install X and a desktop


Add the IP# of the machine running Boinc-Manager, that you will manage the server from, to the remote_hosts.cfg file. Replace 192.168.1.111 with the IP# of your machine that you will control the server from. NOT the IP# of the Ubuntu server.

sudo echo "192.168.1.111" > /etc/boinc-client/remote_hosts.cfg

Restart boinc to read the changed config file.

sudo service boinc-client restart

Get IP# of Ubuntu Server.

ifconfig

poppageek@ghostN ~ $ ifconfig
eth0 Link encap:Ethernet HWaddr 50:e5:49:92:d2:0a
inet addr:192.168.1.101 Bcast:192.168.1.255 Mask:255.255.255.0
inet6 addr: fe80::52e5:49ff:fe92:d3a0/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:465915 errors:36 dropped:0 overruns:36 frame:0
TX packets:685327 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:258852840 (258.8 MB) TX bytes:818445853 (818.4 MB)

lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
UP LOOPBACK RUNNING MTU:65536 Metric:1
RX packets:1394426 errors:0 dropped:0 overruns:0 frame:0
TX packets:1394426 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:10258786746 (10.2 GB) TX bytes:10258786746 (10.2 GB)

wlan0 Link encap:Ethernet HWaddr e4:d5:3d:37:8e:58
inet addr:192.168.1.127 Bcast:192.168.1.255 Mask:255.255.255.0
inet6 addr: fe80::e6d5:3dff:fe73:e885/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:280416 errors:0 dropped:0 overruns:0 frame:0
TX packets:113209 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:277600682 (277.6 MB) TX bytes:15356507 (15.3 MB)

eth0 is the ethernet, wired, network interface. A wireless will be wlan0 or wlan1. Get the IP for which ever you are using.

Add World Community Grid as a project.

At this point choose whether to continue using command line or skip ahead to section below: "From your desktop PC using Boinc Manager:"

From Command Line from machine running Boinc Manager:

Go to the machine you will manage the server from and ssh into the server. For Windows download and install PuTTY On a Linux machine Open a terminal window, Cntrl + ALT + T. Type in:

Replace the IP# below with IP# of the Ubuntu Server.

ssh 192.168.1.101

Enter username and password of the server.

Find your WCG Account Key.

Change username to your WCG account name and password

boinccmd --lookup_account http://www.worldcommunitygrid.org yourusername password

poppageek@ghostN /etc/ssh $ boinccmd --lookup_account http://www.worldcommunitygrid.org poppageek notmypw
status: Success
poll status: operation in progress
account key: 8777c3e6b776c232ao5ea555790o6b32

Use Copy and Paste to copy the account key in the output and paste at the end of the line to attach to project.

boinccmd --host localhost --project_attach http://www.worldcommunitygrid.org 8777c3e6b776c232ao5ea555790o6b32

From your desktop PC using Boinc Manager:

Advanced > Select Computer. enter IP# of Ubuntu Server



Tools > Add Project




Select New or Existing. Enter username and password.





When you are done and want Boinc Manager to display the local machines client: Advanced > Select Computer. enter localhost or the IP# of the desktop.


Listing of boinccmd commands

boinccmd man page

To display boinc man page from command line

man boinccmd

Please advise of errors.

[PoppaGeek]

1. dual quad core opteron 8356 with 4 gigs ram Ubuntu server 64bit 3.2.0-23-generic kernel RAC 3616
2. dual quad core opteron 8356 with 2 gigs ram Ubuntu server 64bit 3.5.0-43-generic kernel RAC 3343

1. Boinc apps run with PRiority of 20 NIce 0
2. Boinc apps run with PRiority of 30 NIce 10

1. shows 2.5 gigs ram free
2. shows .5 gigs ram free

[NKrader]

[D_A]

Try this:

sudo echo "options ath9k nohwcrypt=1" > /etc/modprobe.d/ath9k.conf

and reboot

[D_A]

This may or may not help:

http://www.linuxplained.com/how-to-f...se-pangolin/2/

[NKrader]

should be getting the chassis tonight for this rig so it will be easier to play with as i wont have to setup a cardboard rig everytime I want to try to get it working.

lets hope your comment helps d_a

if not, i reinstalled last night with ssh server (i think)

[PoppaGeek]

should be getting the chassis tonight for this rig so it will be easier to play with as i wont have to setup a cardboard rig everytime I want to try to get it working.

lets hope your comment helps d_a

if not i reinstalled last night with ssh server (i think)

you'll know soon enough

[PoppaGeek]

1. dual quad core opteron 8356 with 4 gigs ram Ubuntu server 64bit 3.2.0-23-generic kernel RAC 3616
2. dual quad core opteron 8356 with 2 gigs ram Ubuntu server 64bit 3.5.0-43-generic kernel RAC 3343

1. Boinc apps run with PRiority of 20 NIce 0
2. Boinc apps run with PRiority of 30 NIce 10

1. shows 2.5 gigs ram free
2. shows .5 gigs ram free

Both machines have <no_priority_change>1</no_priority_change> in the cc_config.xml and SCHEDULE="0" in /etc/default/boinc-client. Yet #2 keeps starting new apps with priority 39 and nice 19. I even tried setting nice and priority values for boinc in the /etc/security/limits.conf.

I can put renice in crontab but I would like to know what keeps setting the PRiority and NIce values on that machine differently than the other.

Have tried many searches and cannot find anything.

[OldChap]

Depends on the project.

MCM1 follows the flag rules as does, I believe, FAAH.

FAHV and CEP, on the other hand, do not. Even though you may set it or reset it using cron, each time it checkpoints it changes back to the project determined settings.

I believe the above to be accurate based on observation and discussion with Rob at WCG

https://secure.worldcommunitygrid.or...ffset,0#439147

[PoppaGeek]

Depends on the project.

MCM1 follows the flag rules as does, I believe, FAAH.

FAHV and CEP, on the other hand, do not. Even though you may set it or reset it using cron, each time it checkpoints it changes back to the project determined settings.

I believe the above to be accurate based on observation and discussion with Rob at WCG

FAHV is one doing it and I THINK a CEP2 did last night. That info saves me alot of additional looking.

I found the variables in the init script for boinc-client that sets priority and nice but have yet to figure out what sets the variables. Never was very good at shell scripting and it's been a loooong time.

POLICY="$1"
PRIO="$2"
NICE="$3"
shift 3
for i in "$@"; do
chrt -p $POLICY $PRIO $i || return
# increasing compatibility with older versions of renice
# in reaction to bug report #600134
renice $NICE -p $i > /dev/null || renice -n $NICE -p $i
done

[OldChap]

added link to mine above whilst you were posting.

If you run just MCM1 you can let the kernel do the scheduling and then watch the processor lost time dwindle to V small.

On FAHV.....No matter what I tried, and that included a cron job every minute, on this rig I cannot get realtime and cpu time anywhere near each other. Did not try under 1 min

[PoppaGeek]

Good thread. Let me know if you take this up with Berkely.