Setting up a Sage server

by Jason Grout

I recently set up a Sage server, and here are very rough notes of what I did. Please email the sage-support email list if you have any questions. Please feel free to make these notes better.

Why set up a server?

I used http://www.sagenb.org for a semester. I then set up a campus server and used that for a semester. Here are the advantages to having a campus server that I found (in no particular order):

  1. At various times, we would experience slowdowns in sagenb.org (it always seemed to happen during a class presentation!). We attributed this to our internet connection, the server being overloaded, or both. A dedicated campus server takes care of both of these issues (provided you budget enough hardware).

  2. I am not comfortable with asking students to publish worksheets on the public server. There are some people that would not be comfortable with having the students do any work on an external server. A campus server can be firewalled to only be accessible from on campus.
  3. While the sagenb.org administrators do a fantastic job, there are no guarantees about service, uptime, backups, etc. Having a campus server allows us to control these things (especially backups!).

  4. We had no control over when upgrades were performed. It was possible that a feature of Sage that we were using would change in a next release. If so, having a campus server allows me to fix a version for the semester and stick with it (or just do upgrades if it won't negatively affect the students' experience), so that the experience is consistent for the semester. On the other hand, there were several features that I contributed to Sage to help the students, and I could apply those immediately on a campus server, instead of having to wait until the next version of Sage.

In our case, we found an old server that was several years old sitting unused. We put 16GB of RAM in it to handle the load we would like to handle and that was our only cost.

Hardware Requirements

RAM

The biggest bottleneck seems to be RAM; that will determine how many simultaneous users you can have. Some anecdotal evidence indicates that, with the 'new' notebook in Sage 5.2 or later, classes of 60-70 take around 12-14GB of RAM (assuming most students have several worksheets open simultaneously). The key to calculating RAM usage is to estimate the number of worksheets open simultaneously, and get an estimate of the sorts of computations people are running. For example, doing integration starts up maxima, which takes a lot of memory.

Install the server

I started with a fresh copy of Ubuntu 10.04, with a working Sage compiled from source (which means I had to install some extra packages so that Sage compiles and runs; see the Sage README. In particular, I needed to install ).

sudo apt-get install libssl-dev

Also, I had to install the following so that Sage would build (see the README.txt file that comes with Sage):

sudo apt-get install build-essential
sudo apt-get install gfortran

I installed these packages to provide extra capabilities in Sage:

sudo apt-get install graphicsmagick-imagemagick-compat # could also install imagemagick instead
sudo apt-get install libpango1.0-dev libcairo-dev # to pick up dependencies for R to enable png support
sudo apt-get install texlive-full # could probably install a subset of texlive instead
sudo apt-get install libreadline-dev # for qepcad

(I think the R png capabilities line is right. If it's not, you may have to also install the xorg-dev, libpng-dev, and/or r-base packages)

I installed this to help me manage the server better:

sudo apt-get install unattended-upgrades
sudo vim /etc/apt/apt.conf.d/50unattended-upgrades # edit unattended upgrades config; see https://help.ubuntu.com/10.04/serverguide/C/automatic-updates.html
sudo apt-get install mercurial

sudo apt-get install haproxy

# this config needs haproxy-1.1.28 or haproxy-1.2.1

global
        log 127.0.0.1   local0
        log 127.0.0.1   local1 notice
        #log loghost    local0 info
        maxconn 4096
        #chroot /usr/share/haproxy
        user haproxy
        group haproxy
        daemon
        #debug
        #quiet

defaults
        log     global
        mode    http
        option  httplog
        option  dontlognull
        retries 3
        option redispatch
        maxconn 2000
        contimeout      5000
        clitimeout      50000
        srvtimeout      50000

listen  sageserver 0.0.0.0:80
        server  sageserver localhost:8000 check 

sudo /etc/init.d/haproxy start

sudo addgroup sageuser
sudo adduser --disabled-password sageserver
for i in $(seq 0 9); do
 sudo adduser --disabled-password --ingroup sageuser sage$i
done

account  required     pam_access.so nodefgroup

-:(sageuser):ALL EXCEPT localhost
-:sageserver:ALL EXCEPT <backup>

sudo -u sageserver -i ssh-keygen -t dsa
for i in $(seq 0 9); do
 sudo cat ~sageserver/.ssh/id_dsa.pub | sudo -u sage$i -i "umask 077; test -d .ssh || mkdir .ssh ; cat >> .ssh/authorized_keys "
done

sudo -u sageserver ssh sage0@localhost echo Done

 #!/bin/sh

~/sage/sage -c "notebook(interface='localhost', directory='./sage_notebook.sagenb',port=8000, accounts=True, timeout=3600, server_pool=['sage%d@localhost'%i for i in range(10)], ulimit='-u 100 -t 36000 -v 500000', open_viewer=False)" 

Note that I explicitly set the notebook directory, so that the directory doesn't default to live inside of ~/.sage. This is so that the normal procedure of making ~/.sage not publicly accessible doesn't also cut off access to each cell's DATA directory.

   cd ~sageserver
   ln -s sage-4.3.2 sage

for i in 'biopython' 'cbc' 'glpk' 'graphviz' 'pyx' 'dot2tex' 'fricas'; do
    sudo -u sageserver ~sageserver/sage/sage -i $i;
done;

To start the sage server, do the following. Note that since I am using sudo to run commands as sageserver, instead of logging in as sageserver, I have to do the script /dev/null trick to get screen to work.

sudo su -l sageserver
script /dev/null
screen ./startnotebook

I also added this to ~/sage/sage to control process limits:

if [[ `whoami` = sage* ]]; then
   echo "User " `whoami`
   ulimit -v 1500000 -u 300 -n 128 -t 1800
fi

Additional Notes

sudo apt-get install python-software-properties
sudo add-apt-repository ppa:webupd8team/java
sudo apt-get update
sudo apt-get install oracle-java7-installer

Alternative port forwarding arrangements

Apart from Apache, it may also be possible to use iptables or some other NAT firewall to do the port forwarding. In that case you should of course run the sage notebook with secure=True, because a firewall will not be providing an SSL layer for you. The command to forward, for instance, port 443 to port 8000 on network interface eth0 (you should check if that is the interface on which the connections come in) is

iptables -t nat -A PREROUTING -i eth0 -p tcp --dport 443 -j REDIRECT --to-port 8000

You should check with your own system to see how to make such a rule permanent. If you want check which nat rules are in effect type

iptables -t nat -L

Other Instructions

You can find another set of instructions at JustEnoughSageServer.

SageServer (last edited 2013-12-04 18:44:19 by kcrisman)