source: trunk/server/doc/install-howto.sh @ 1866

Last change on this file since 1866 was 1704, checked in by ezyang, 13 years ago
Info about adding servers to SSH config.
File size: 16.5 KB
RevLine 
[1058]1# This document is a how-to for installing a Fedora scripts.mit.edu server.
[1693]2# It is semi-vaguely in the form of a shell script, but is not really
3# runnable as it stands.
[181]4
[1063]5set -e -x
6
[1693]7# Some commands should be run as the scripts-build user, not root.
[1063]8
[1693]9alias asbuild="sudo -u scripts-build"
[1063]10
[1693]11# Old versions of this install document advised setting
12# NSS_NONLOCAL_IGNORE=1 anytime you're setting up anything, e.g. using
13# yum, warning that useradd will query LDAP in a stupid way that makes
14# it hang forever.  As of Fedora 13, this does not seem to be a problem,
15# so it's been removed from the instructions.  If an install is hanging,
16# though, try adding NSS_NONLOCAL_IGNORE.
[1063]17
[1693]18# This is actually just "pick an active scripts server".  It can't be
19# scripts.mit.edu because our networking config points that domain
20# at localhost, and if our server is not setup at that point things
21# will break.
22source_server="cats-whiskers.mit.edu"
[1259]23
[1693]24# 'branch' is the current svn branch you are on.  You want to
25# use trunk if your just installing a new server, and branches/fcXX-dev
26# if your preparing a server on a new Fedora release.
27branch="trunk"
[1063]28
[1693]29# 'server' is the public hostname of your server, for SCP'ing files
30# to and from.
31server=YOUR-SERVER-NAME-HERE
[1063]32
[1693]33# Start with a Scripts kickstarted install of Fedora (install-fedora)
[861]34
[1693]35# Take updates, reboot if there's a kernel update.
[1698]36    yum update -y
[861]37
[1693]38# Get rid of network manager
39    yum remove NetworkManager
[1381]40
[1693]41# Copy over root's dotfiles from one of the other machines.
42# Perhaps a useful change is to remove the default aliases
43    cd /root
44    ls -l .bashrc
45    ls -l .ldapvirc
46    ls -l .screenrc
47    ls -l .ssh
48    ls -l .vimrc
49    ls -l .k5login
50    # Trying to scp from server to server won't work, as scp
51    # will attempt to negotiate a server-to-server connection.
52    # Instead, scp to your trusted machine as a temporary file,
53    # and then push to the other server
54scp -r root@$source_server:~/{.bashrc,.ldapvirc,.screenrc,.ssh,.vimrc,.k5login} .
55scp -r {.bashrc,.ldapvirc,.screenrc,.ssh,.vimrc,.k5login} root@$server:~
[1063]56
[1693]57# Install the initial set of credentials (to get Kerberized logins once
58# krb5 is installed).  Otherwise, SCP'ing things in will be annoying.
59#   o You probably installed the machine keytab long ago
60    ls -l /etc/krb5.keytab
61#     Use ktutil to combine the host/scripts.mit.edu and
62#     host/scripts-vhosts.mit.edu keys with host/this-server.mit.edu in
63#     the keytab.  Do not use 'k5srvutil change' on the combined keytab
64#     or you'll break the other servers. (real servers only).  Be
65#     careful about writing out the keytab: if you write it to an
66#     existing file the keys will just get appended.  The correct
67#     credential list should look like:
68#       ktutil:  l
69#       slot KVNO Principal
70#       ---- ---- ---------------------------------------------------------------------
71#          1    5 host/old-faithful.mit.edu@ATHENA.MIT.EDU
72#          2    3 host/scripts-vhosts.mit.edu@ATHENA.MIT.EDU
73#          3    2      host/scripts.mit.edu@ATHENA.MIT.EDU
74#   o Replace the ssh host keys with the ones common to all scripts servers (real servers only)
75    ls -l /etc/ssh/*key*
76#     You can do that with:
77scp root@$source_server:/etc/ssh/*key* .
78scp *key* root@$server:/etc/ssh/
79    service sshd reload
[1063]80
81# Check out the scripts /etc configuration
[1693]82    # backslash to make us not use the alias
[1063]83    cd /root
84    \cp -a etc /
[1693]85    chmod 0440 /etc/sudoers
[785]86
[1696]87# If this is the first time you've installed this hostname, you will
88# need to update a bunch of files to add support for it. These include:
89#   o Adding all aliases to /etc/httpd/conf.d/scripts-vhost-names.conf
90#     (usually this is hostname, hostname.mit.edu, h-n, h-n.mit.edu,
91#     scriptsN, scriptsN.mit.edu, and the IP address.)
92#   o Adding routing rules for the static IP in
93#     /etc/sysconfig/network-scripts/route-eth1
94#   o Adding the IP address to the hosts file (same hosts as for
95#     scripts-vhost-names)
[1704]96#   o Update SSH config at
97#       - server/fedora/config/etc/ssh/shosts.equiv
98#       - server/fedora/config/etc/ssh/ssh_known_hosts
99#       - server/fedora/config/etc/ssh/sshd_config : DenyUsers
100#     (the last part is critical to ensure that rooting one server
101#     doesn't give you root to all the other servers)
[1696]102#   o Put the hostname information in LDAP so SVN and Git work
103#   o Set up Nagios monitoring on sipb-noc for the host
104#   o Set up the host as in the pool on r-b/r-b /etc/heartbeat/ldirectord.cf
105    XXX TODO COMMANDS
106
[1381]107# NOTE: You will have just lost DNS resolution and the ability
108# to do password SSH in.  If you managed to botch this step without
109# having named setup, you can do a quick fix by frobbing /etc/resolv.conf
110# with a non 127.0.0.1 address for the DNS server.  Be sure to revert it once
111# you have named.
[1259]112
[1693]113# NOTE: You can get password SSH back by editing /etc/ssh/sshd_config (allow
114# password auth) and /etc/pam.d/sshd (comment out the first three auth
115# lines).  However, you should have the Kerberos credentials in place
116# so as soon as you install the full set of Scripts packages, you'll get
117# Kerberized logins.
[1259]118
[1693]119# Make sure network is working.  If this is a new server name, you'll
120# need to add it to /etc/hosts and
121# /etc/sysconfig/network-scripts/route-eth1.  Kickstart should have
122# configured eth0 and eth1 correctly; use service network restart
123# to add the new routes in route-eth1.
124    service network restart
125    route
126    ifconfig
127    cat /etc/hosts
128    cat /etc/sysconfig/network-scripts/route-eth1
[1259]129
[1693]130# This is the point at which you should start updating scriptsified
131# packages for a new Fedora release.  Consult 'upgrade-tips' for more
132# information.
133    yum install -y scripts-base
134    # Some of these packages are naughty and clobber some of our files
135    cd /etc
136    svn revert resolv.conf hosts sysconfig/openafs
[1259]137
[1058]138# Replace rsyslog with syslog-ng by doing:
[1259]139    rpm -e --nodeps rsyslog
[1693]140    yum install -y syslog-ng
[1259]141    chkconfig syslog-ng on
[861]142
[1693]143# Fix the openafs /usr/vice/etc <-> /etc/openafs mapping.
[1259]144    echo "/afs:/usr/vice/cache:10000000" > /usr/vice/etc/cacheinfo
[1693]145    echo "athena.mit.edu" > /usr/vice/etc/ThisCell
[788]146
[1693]147# [TEST SERVER] If you're installing a test server, this needs to be
148# much smaller; the max filesize on XVM is 10GB.  Pick something like
149# 500000. Also, some of the AFS parameters are kind of retarded (and if
150# you're low on disk space, will actually exhaust our inodes).  Edit
151# these parameters in /etc/sysconfig/openafs
[861]152
[1693]153# Test that zephyr is working
154    chkconfig zhm on
155    service zhm start
156    echo 'Test!' | zwrite -d -c scripts -i test
[1178]157
[1058]158# Install the full list of RPMs that users expect to be on the
[1259]159# scripts.mit.edu servers.
160rpm -qa --queryformat "%{Name}.%{Arch}\n" | sort > packages.txt
161# arrange for packages.txt to be passed to the server, then run:
[1693]162# --skip-broken will (usually) prevent you from having to sit through
163# several minutes of dependency resolution until it decides that
164# it can't install /one/ package.
165    yum install -y --skip-broken $(cat packages.txt)
[1259]166
[1698]167# Make sure sendmail isn't installed
168    yum remove sendmail
169
[1190]170# Check which packages are installed on your new server that are not
171# in the snapshot, and remove ones that aren't needed for some reason
172# on the new machine.  Otherwise, aside from bloat, you may end up
173# with undesirable things for security, like sendmail.
[1693]174    rpm -qa --queryformat "%{Name}.%{Arch}\n" | grep -v kernel | sort > newpackages.txt
175    diff -u packages.txt newpackages.txt | grep -v kernel | less
[1382]176    # here's a cute script that removes all extra packages
[1693]177    yum erase -y $(grep -Fxvf packages.txt newpackages.txt)
[1190]178
[1693]179# We need an upstream version of cgi which we've packaged ourselves, but
180# it doesn't work with the haskell-platform package which expects
181# explicit versions.  So temporarily rpm -e the package, and then
182# install it again after you install haskell-platform.  [Note: You
183# probably won't need this in Fedora 15 or something, when the Haskell
184# Platform gets updated.]
185    rpm -e ghc-cgi-devel ghc-cgi
186    yum install -y haskell-platform
187    yumdownloader ghc-cgi
188    yumdownloader ghc-cgi-devel
189    rpm -i ghc-cgi*1.8.1*.rpm
190
[1237]191# Check out the scripts /usr/vice/etc configuration
[1693]192    cd /root/vice
[1237]193    \cp -a etc /usr/vice
194
[1058]195# Install the full list of perl modules that users expect to be on the
[1108]196# scripts.mit.edu servers.
[1693]197    cd /root
[1259]198    export PERL_MM_USE_DEFAULT=1
[1693]199    cpan # this is interactive, enter the next two lines
[1259]200        o conf prerequisites_policy follow
201        o conf commit
202# on a reference server
203perldoc -u perllocal | grep head2 | cut -f 3 -d '<' | cut -f 1 -d '|' | sort -u | perl -ne 'chomp; print "notest install $_\n" if system("rpm -q --whatprovides \"perl($_)\" >/dev/null 2>/dev/null")' > perl-packages.txt
204# arrange for perl-packages.txt to be transferred to server
205    cat perl-packages.txt | perl -MCPAN -e shell
[812]206
[1058]207# Install the Python eggs and Ruby gems and PEAR/PECL doohickeys that are on
208# the other scripts.mit.edu servers and do not have RPMs.
[1259]209# The general mode of operation will be to run the "list" command
210# on both servers, see what the differences are, check if those diffs
211# are packaged up as rpms, and install them (rpm if possible, native otherwise)
[1178]212# - Look at /usr/lib/python2.6/site-packages and
213#           /usr/lib64/python2.6/site-packages for Python eggs and modules.
[1259]214#   There will be a lot of gunk that was installed from packages;
[1693]215#   easy-install.pth in /usr/lib/ will tell you what was easy_installed.
[1178]216#   First use 'yum search' to see if the relevant package is now available
217#   as an RPM, and install that if it is.  If not, then use easy_install.
[1432]218#   Pass -Z to easy_install to install them unzipped, as some zipped eggs
219#   want to be able to write to ~/.python-eggs.  (Also makes sourcediving
220#   easier.)
[1698]221cat /usr/lib/python2.6/site-packages/easy-install.pth | grep "^./" | cut -c3- | cut -f1 -d- . egg.txt
222    cat egg.txt | xargs easy_install -Z
[1058]223# - Look at `gem list` for Ruby gems.
[1178]224#   Again, use 'yum search' and prefer RPMs, but failing that, 'gem install'.
[1259]225#       ezyang: rspec-rails depends on rspec, and will override the Yum
226#       package, so... don't use that RPM yet
[1693]227gem list --no-version > gem.txt
228    gem install $(gem list --no-version | grep -Fxvf - gem.txt)
[1058]229# - Look at `pear list` for Pear fruits (or whatever they're called).
[1178]230#   Yet again, 'yum search' for RPMs before resorting to 'pear install'.  Note
231#   that for things in the beta repo, you'll need 'pear install package-beta'.
[1259]232#   (you might get complaints about the php_scripts module; ignore them)
[1693]233pear list | tail -n +4 | cut -f 1 -d " " > pear.txt
234    pear config-set preferred_state beta
235    pear channel-update pear.php.net
236    pear install $(pear list | tail -n +4 | cut -f 1 -d " " | grep -Fxvf - pear.txt)
[1190]237# - Look at `pecl list` for PECL things.  'yum search', and if you must,
[1462]238#   'pecl install' needed items. If it doesn't work, try 'pear install
[1544]239#   pecl/foo' or 'pecl install foo-beta' or those two combined.
[1693]240pecl list | tail -n +4 | cut -f 1 -d " " > pecl.txt
241    pecl install --nodeps $(pecl list | tail -n +4 | cut -f 1 -d " " | grep -Fxvf - pecl.txt)
[785]242
[1259]243# Setup some Python config
244    echo 'import site, os.path; site.addsitedir(os.path.expanduser("~/lib/python2.6/site-packages"))' > /usr/lib/python2.6/site-packages/00scripts-home.pth
[812]245
[1693]246# Install the credentials.  There are a lot of things to remember here.
247# Be sure to make sure the permissions match up (ls -l on an existing
248# server!).
249scp root@$source_server:{/etc/{sql-mit-edu.cfg.php,daemon.keytab,pki/tls/private/scripts.key,signup-ldap-pw,whoisd-password},/home/logview/.k5login} .
250scp daemon.keytab signup-ldap-pw whoisd-password sql-mit-edu.cfg.php root@$server:/etc
251scp scripts.key root@$server:/etc/pki/tls/private
252scp .k5login root@$server:/home/logview
253    chown afsagent:afsagent /etc/daemon.keytab
254#   o The daemon.scripts keytab (will be daemon.scripts-test for test)
[1259]255    ls -l /etc/daemon.keytab
256#   o The SSL cert private key (real servers only)
[1693]257    ls -l /etc/pki/tls/private/scripts.key
[1259]258#   o The LDAP password for the signup process (real servers only)
[1693]259    ls -l /etc/signup-ldap-pw
[1320]260#   o The whoisd password (real servers only)
[1693]261    ls -l /etc/whoisd-password
[1259]262#   o Make sure logview's .k5login is correct (real servers only)
[1693]263    cat /home/logview/.k5login
[387]264
[1693]265# Spin up OpenAFS.  This will fail if there's been a new kernel since
266# when you last tried.  In that case, you can hold on till later to
267# start OpenAFS.  This will take a little bit of time;
268    service openafs-client start
269
270# Check that fs sysname is correct.  You should see, among others,
271# 'amd64_fedoraX_scripts' (vary X) and 'scripts'. If it's not, you
272# probably did a distro upgrade and should update /etc/sysconfig/openafs.
273    fs sysname
274
275# [TEST SERVER] If you are setting up a test server, pay attention to
[1058]276# /etc/sysconfig/network-scripts and do not bind scripts' IP address.
[1259]277# You will also need to modify:
278#   o /etc/ldap.conf
279#       add: host scripts.mit.edu
280#   o /etc/nss-ldapd.conf
281#       replace: uri *****
282#       with: uri ldap://scripts.mit.edu/
283#   o /etc/openldap/ldap.conf
284#       add: URI ldap://scripts.mit.edu/
285#            BASE dc=scripts,dc=mit,dc=edu
286#   o /etc/httpd/conf.d/vhost_ldap.conf
287#       replace: VhostLDAPUrl ****
288#       with: VhostLDAPUrl "ldap://scripts.mit.edu/ou=VirtualHosts,dc=scripts,dc=mit,dc=edu"
[1450]289#   o /etc/postfix/virtual-alias-{domains,maps}-ldap.cf
290#       replace: server_host *****
291#       with: server_host = ldap://scripts.mit.edu
[1259]292# to use scripts.mit.edu instead of localhost.
293# XXX: someone should write sed scripts to do this
[562]294
[1693]295# [TEST SERVER] If you are setting up a test server, afsagent's cronjob
296# will attempt to be renewing with the wrong credentials
297# (daemon.scripts). Change this:
[1259]298    vim /home/afsagent/renew # replace all mentions of daemon.scripts.mit.edu
299
[1693]300# Set up replication (see ./install-ldap).
301# You'll need the LDAP keytab for this server: be sure to chown it
302# fedora-ds after you create the fedora-ds user
303    ls -l /etc/dirsrv/keytab
304    cat install-ldap
[785]305
[1058]306# Make the services dirsrv, nslcd, nscd, postfix, and httpd start at
307# boot. Run chkconfig to make sure the set of services to be run is
308# correct.
[1693]309    service nslcd start
310    service nscd start
311    service postfix start
312    service httpd start
[1259]313    chkconfig dirsrv on
314    chkconfig nslcd on
315    chkconfig nscd on
316    chkconfig postfix on
317    chkconfig httpd on
[818]318
[1687]319# nrpe is required for nagios alerts
320    chkconfig nrpe on
321
[1693]322# Check sql user credentials (needs to be done after LDAP is setup)
323    chown sql /etc/sql-mit-edu.cfg.php
324
[1259]325# Postfix doesn't actually deliver mail; fix this
326    cd /etc/postfix
327    postmap virtual
[1178]328
[1451]329# Munin might not be monitoring packages that were installed after it
330    munin-node-configure --suggest --shell | sh
331
[1058]332# Run fmtutil-sys --all, which does something that makes TeX work.
[1693]333# (Note: this errors on XeTeX which is ok.)
[1259]334    fmtutil-sys --all
[803]335
[1058]336# Ensure that PHP isn't broken:
[1259]337    mkdir /tmp/sessions
338    chmod 01777 /tmp/sessions
[1699]339    # XXX: this seems to get deleted if tmp gets cleaned up, so we
340    # might need something a little better (maybe init script.)
[954]341
[1693]342# Ensure fcgid isn't broken (should be 755)
343    ls -ld /var/run/mod_fcgid
[1259]344
345# Fix etc by making sure none of our config files got overwritten
346    cd /etc
[1693]347    svn status -q
348    # Some usual candidates for clobbering include nsswitch.conf and
349    # sysconfig/openafs
[1259]350
[1382]351# ThisCell got clobbered, replace it with athena.mit.edu
352    echo "athena.mit.edu" > /usr/vice/etc/ThisCell
353
[1058]354# Reboot the machine to restore a consistent state, in case you
[1693]355# changed anything. (Note: Starting kdump fails (this is ok))
[875]356
[1693]357# [OPTIONAL] Your machine's hostname is baked in at install time;
358# in the rare case you need to change it: it appears to be in:
[1259]359#   o /etc/sysconfig/network
360#   o your lvm thingies; probably don't need to edit
[1382]361
[1693]362# [TEST SERVER] More stuff for test servers
[1382]363#   - You need a self-signed SSL cert.  Generate with:
364    openssl req -new -x509 -keyout /etc/pki/tls/private/scripts.key -out /etc/pki/tls/certs/scripts.cert -nodes
365#     Also make /etc/pki/tls/certs/ca.pem match up
366#   - Make (/etc/aliases) root mail go to /dev/null, so we don't spam people
367#   - Edit /etc/httpd/conf.d/scripts-vhost-names.conf to have scripts-fX-test.xvm.mit.edu
368#     be an accepted vhost name
369#   - Look at the old test server and see what config changes are floating around
[1693]370
371# XXX: our SVN checkout should be updated to use scripts.mit.edu
372# (repository and etc) once serving actually works.
373    cd /etc
374    svn switch --relocate svn://$source_server/ svn://scripts.mit.edu/
375    cd /usr/vice/etc
376    svn switch --relocate svn://$source_server/ svn://scripts.mit.edu/
377    cd /srv/repository
378    asbuild svn switch --relocate svn://$source_server/ svn://scripts.mit.edu/
379    asbuild svn up # verify scripts.mit.edu works
Note: See TracBrowser for help on using the repository browser.