Merge branch '19437-docker-watchdog'
[arvados.git] / tools / salt-install / provision.sh
1 #!/bin/bash
2
3 # Copyright (C) The Arvados Authors. All rights reserved.
4 #
5 # SPDX-License-Identifier: CC-BY-SA-3.0
6
7 # If you want to test arvados in a single host, you can run this script, which
8 # will install it using salt masterless
9 # This script is run by the Vagrant file when you run it with
10 #
11 # vagrant up
12
13 set -o pipefail
14
15 # capture the directory that the script is running from
16 SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
17
18 usage() {
19   echo >&2
20   echo >&2 "Usage: ${0} [-h] [-h]"
21   echo >&2
22   echo >&2 "${0} options:"
23   echo >&2 "  -d, --debug                                 Run salt installation in debug mode"
24   echo >&2 "  -c <local.params>, --config <local.params>  Path to the local.params config file"
25   echo >&2 "  -t, --test                                  Test installation running a CWL workflow"
26   echo >&2 "  -r, --roles                                 List of Arvados roles to apply to the host, comma separated"
27   echo >&2 "                                              Possible values are:"
28   echo >&2 "                                                api"
29   echo >&2 "                                                controller"
30   echo >&2 "                                                dispatcher"
31   echo >&2 "                                                keepproxy"
32   echo >&2 "                                                keepbalance"
33   echo >&2 "                                                keepstore"
34   echo >&2 "                                                keepweb"
35   echo >&2 "                                                shell"
36   echo >&2 "                                                webshell"
37   echo >&2 "                                                websocket"
38   echo >&2 "                                                workbench"
39   echo >&2 "                                                workbench2"
40   echo >&2 "                                              Defaults to applying them all"
41   echo >&2 "  -h, --help                                  Display this help and exit"
42   echo >&2 "  --dump-config <dest_dir>                    Dumps the pillars and states to a directory"
43   echo >&2 "                                              This parameter does not perform any installation at all. It's"
44   echo >&2 "                                              intended to give you a parsed set of configuration files so"
45   echo >&2 "                                              you can inspect them or use them in you Saltstack infrastructure."
46   echo >&2 "                                              It"
47   echo >&2 "                                                - parses the pillar and states templates,"
48   echo >&2 "                                                - downloads the helper formulas with their desired versions,"
49   echo >&2 "                                                - prepares the 'top.sls' files both for pillars and states"
50   echo >&2 "                                                  for the selected role(s)"
51   echo >&2 "                                                - writes the resulting files into <dest_dir>"
52   echo >&2 "  -v, --vagrant                               Run in vagrant and use the /vagrant shared dir"
53   echo >&2 "  --development                               Run in dev mode, using snakeoil certs"
54   echo >&2
55 }
56
57 arguments() {
58   # NOTE: This requires GNU getopt (part of the util-linux package on Debian-based distros).
59   if ! which getopt > /dev/null; then
60     echo >&2 "GNU getopt is required to run this script. Please install it and re-reun it"
61     exit 1
62   fi
63
64   TEMP=$(getopt -o c:dhp:r:tv \
65     --long config:,debug,development,dump-config:,help,roles:,test,vagrant \
66     -n "${0}" -- "${@}")
67
68   if [ ${?} != 0 ];
69     then echo "Please check the parameters you entered and re-run again"
70     exit 1
71   fi
72   # Note the quotes around `$TEMP': they are essential!
73   eval set -- "$TEMP"
74
75   while [ ${#} -ge 1 ]; do
76     case ${1} in
77       -c | --config)
78         CONFIG_FILE=${2}
79         shift 2
80         ;;
81       -d | --debug)
82         LOG_LEVEL="debug"
83         shift
84         set -x
85         ;;
86       --dump-config)
87         if [[ ${2} = /* ]]; then
88           DUMP_SALT_CONFIG_DIR=${2}
89         else
90           DUMP_SALT_CONFIG_DIR=${PWD}/${2}
91         fi
92         ## states
93         S_DIR="${DUMP_SALT_CONFIG_DIR}/salt"
94         ## formulas
95         F_DIR="${DUMP_SALT_CONFIG_DIR}/formulas"
96         ## pillars
97         P_DIR="${DUMP_SALT_CONFIG_DIR}/pillars"
98         ## tests
99         T_DIR="${DUMP_SALT_CONFIG_DIR}/tests"
100         DUMP_CONFIG="yes"
101         shift 2
102         ;;
103       --development)
104         DEV_MODE="yes"
105         shift 1
106         ;;
107       -r | --roles)
108         for i in ${2//,/ }
109           do
110             # Verify the role exists
111             if [[ ! "database,api,controller,keepstore,websocket,keepweb,workbench2,webshell,keepbalance,keepproxy,shell,workbench,dispatcher" == *"$i"* ]]; then
112               echo "The role '${i}' is not a valid role"
113               usage
114               exit 1
115             fi
116             ROLES="${ROLES} ${i}"
117           done
118           shift 2
119         ;;
120       -t | --test)
121         TEST="yes"
122         shift
123         ;;
124       -v | --vagrant)
125         VAGRANT="yes"
126         shift
127         ;;
128       --)
129         shift
130         break
131         ;;
132       *)
133         usage
134         exit 1
135         ;;
136     esac
137   done
138 }
139
140 copy_custom_cert() {
141   cert_dir=${1}
142   cert_name=${2}
143
144   mkdir -p /srv/salt/certs
145
146   if [ -f ${cert_dir}/${cert_name}.crt ]; then
147     cp -v ${cert_dir}/${cert_name}.crt /srv/salt/certs/arvados-${cert_name}.pem
148   else
149     echo "${cert_dir}/${cert_name}.crt does not exist. Exiting"
150     exit 1
151   fi
152   if [ -f ${cert_dir}/${cert_name}.key ]; then
153     cp -v ${cert_dir}/${cert_name}.key /srv/salt/certs/arvados-${cert_name}.key
154   else
155     echo "${cert_dir}/${cert_name}.key does not exist. Exiting"
156     exit 1
157   fi
158 }
159
160 DEV_MODE="no"
161 CONFIG_FILE="${SCRIPT_DIR}/local.params"
162 CONFIG_DIR="local_config_dir"
163 DUMP_CONFIG="no"
164 LOG_LEVEL="info"
165 CONTROLLER_EXT_SSL_PORT=443
166 TESTS_DIR="tests"
167
168 NGINX_INSTALL_SOURCE="install_from_repo"
169
170 CLUSTER=""
171 DOMAIN=""
172
173 # Hostnames/IPs used for single-host deploys
174 IP_INT="127.0.1.1"
175
176 # Initial user setup
177 INITIAL_USER=""
178 INITIAL_USER_EMAIL=""
179 INITIAL_USER_PASSWORD=""
180
181 CONTROLLER_EXT_SSL_PORT=8000
182 KEEP_EXT_SSL_PORT=25101
183 # Both for collections and downloads
184 KEEPWEB_EXT_SSL_PORT=9002
185 WEBSHELL_EXT_SSL_PORT=4202
186 WEBSOCKET_EXT_SSL_PORT=8002
187 WORKBENCH1_EXT_SSL_PORT=443
188 WORKBENCH2_EXT_SSL_PORT=3001
189
190 SSL_MODE="self-signed"
191 USE_LETSENCRYPT_ROUTE53="no"
192 CUSTOM_CERTS_DIR="${SCRIPT_DIR}/local_config_dir/certs"
193
194 ## These are ARVADOS-related parameters
195 # For a stable release, change RELEASE "production" and VERSION to the
196 # package version (including the iteration, e.g. X.Y.Z-1) of the
197 # release.
198 # The "local.params.example.*" files already set "RELEASE=production"
199 # to deploy  production-ready packages
200 RELEASE="development"
201 VERSION="latest"
202
203 # These are arvados-formula-related parameters
204 # An arvados-formula tag. For a stable release, this should be a
205 # branch name (e.g. X.Y-dev) or tag for the release.
206 # ARVADOS_TAG="2.2.0"
207 # BRANCH="main"
208
209 # We pin the salt version to avoid potential incompatibilities when a new
210 # stable version is released.
211 SALT_VERSION="3004"
212
213 # Other formula versions we depend on
214 POSTGRES_TAG="v0.44.0"
215 NGINX_TAG="v2.8.1"
216 DOCKER_TAG="v2.4.2"
217 LOCALE_TAG="v0.3.4"
218 LETSENCRYPT_TAG="v2.1.0"
219 LOGROTATE_TAG="v0.14.0"
220
221 # Salt's dir
222 DUMP_SALT_CONFIG_DIR=""
223 ## states
224 S_DIR="/srv/salt"
225 ## formulas
226 F_DIR="/srv/formulas"
227 ## pillars
228 P_DIR="/srv/pillars"
229 ## tests
230 T_DIR="/tmp/cluster_tests"
231
232 arguments ${@}
233
234 declare -A NODES
235
236 if [ -s ${CONFIG_FILE} ]; then
237   source ${CONFIG_FILE}
238 else
239   echo >&2 "You don't seem to have a config file with initial values."
240   echo >&2 "Please create a '${CONFIG_FILE}' file as described in"
241   echo >&2 "  * https://doc.arvados.org/install/salt-single-host.html#single_host, or"
242   echo >&2 "  * https://doc.arvados.org/install/salt-multi-host.html#multi_host_multi_hostnames"
243   exit 1
244 fi
245
246 if [ ! -d ${CONFIG_DIR} ]; then
247   echo >&2 "You don't seem to have a config directory with pillars and states."
248   echo >&2 "Please create a '${CONFIG_DIR}' directory (as configured in your '${CONFIG_FILE}'). Please see"
249   echo >&2 "  * https://doc.arvados.org/install/salt-single-host.html#single_host, or"
250   echo >&2 "  * https://doc.arvados.org/install/salt-multi-host.html#multi_host_multi_hostnames"
251   exit 1
252 fi
253
254 if grep -rni 'fixme' ${CONFIG_FILE} ${CONFIG_DIR} ; then
255   echo >&2 "The config file ${CONFIG_FILE} has some parameters that need to be modified."
256   echo >&2 "Please, fix them and re-run the provision script."
257   exit 1
258 fi
259
260 if ! grep -qE '^[[:alnum:]]{5}$' <<<${CLUSTER} ; then
261   echo >&2 "ERROR: <CLUSTER> must be exactly 5 lowercase alphanumeric characters long"
262   echo >&2 "Fix the cluster name in the 'local.params' file and re-run the provision script"
263   exit 1
264 fi
265
266 # Only used in single_host/single_name deploys
267 if [ ! -z "${HOSTNAME_EXT}" ] ; then
268   # We need to add some extra control vars to manage a single certificate vs. multiple
269   USE_SINGLE_HOSTNAME="yes"
270   # Make sure that the value configured as IP_INT is a real IP on the system.
271   # If we don't error out early here when there is a mismatch, the formula will
272   # fail with hard to interpret nginx errors later on.
273   ip addr list |grep "${IP_INT}/" >/dev/null
274   if [[ $? -ne 0 ]]; then
275     echo "Unable to find the IP_INT address '${IP_INT}' on the system, please correct the value in local.params. Exiting..."
276     exit 1
277   fi
278 else
279   USE_SINGLE_HOSTNAME="no"
280   # We set this variable, anyway, so sed lines do not fail and we don't need to add more
281   # conditionals
282   HOSTNAME_EXT="${CLUSTER}.${DOMAIN}"
283 fi
284
285 if [ "${DUMP_CONFIG}" = "yes" ]; then
286   echo "The provision installer will just dump a config under ${DUMP_SALT_CONFIG_DIR} and exit"
287 else
288   # Install a few dependency packages
289   # First, let's figure out the OS we're working on
290   OS_ID=$(grep ^ID= /etc/os-release |cut -f 2 -d=  |cut -f 2 -d \")
291   echo "Detected distro: ${OS_ID}"
292
293   case ${OS_ID} in
294     "centos")
295       echo "WARNING! Disabling SELinux, see https://dev.arvados.org/issues/18019"
296       sed -i 's/SELINUX=enforcing/SELINUX=permissive/g' /etc/sysconfig/selinux
297       setenforce permissive
298       yum install -y  curl git jq
299       ;;
300     "debian"|"ubuntu")
301       # Wait 2 minutes for any apt locks to clear
302       # This option is supported from apt 1.9.1 and ignored in older apt versions.
303       # Cf. https://blog.sinjakli.co.uk/2021/10/25/waiting-for-apt-locks-without-the-hacky-bash-scripts/
304       DEBIAN_FRONTEND=noninteractive apt -o DPkg::Lock::Timeout=120 update
305       DEBIAN_FRONTEND=noninteractive apt install -y curl git jq
306       ;;
307   esac
308
309   if which salt-call; then
310     echo "Salt already installed"
311   else
312     curl -L https://bootstrap.saltstack.com -o /tmp/bootstrap_salt.sh
313     sh /tmp/bootstrap_salt.sh -XdfP -x python3 stable ${SALT_VERSION}
314     /bin/systemctl stop salt-minion.service
315     /bin/systemctl disable salt-minion.service
316   fi
317
318   # Set salt to masterless mode
319   cat > /etc/salt/minion << EOFSM
320 failhard: "True"
321
322 file_client: local
323 file_roots:
324   base:
325     - ${S_DIR}
326     - ${F_DIR}/*
327
328 pillar_roots:
329   base:
330     - ${P_DIR}
331 EOFSM
332 fi
333
334 mkdir -p ${S_DIR} ${F_DIR} ${P_DIR} ${T_DIR}
335
336 # Get the formula and dependencies
337 cd ${F_DIR} || exit 1
338 echo "Cloning formulas"
339 test -d docker && ( cd docker && git fetch ) \
340   || git clone --quiet https://github.com/saltstack-formulas/docker-formula.git ${F_DIR}/docker
341 ( cd docker && git checkout --quiet tags/"${DOCKER_TAG}" )
342
343 echo "...locale"
344 test -d locale && ( cd locale && git fetch ) \
345   || git clone --quiet https://github.com/saltstack-formulas/locale-formula.git ${F_DIR}/locale
346 ( cd locale && git checkout --quiet tags/"${LOCALE_TAG}" )
347
348 echo "...nginx"
349 test -d nginx && ( cd nginx && git fetch ) \
350   || git clone --quiet https://github.com/saltstack-formulas/nginx-formula.git ${F_DIR}/nginx
351 ( cd nginx && git checkout --quiet tags/"${NGINX_TAG}" )
352
353 echo "...postgres"
354 test -d postgres && ( cd postgres && git fetch ) \
355   || git clone --quiet https://github.com/saltstack-formulas/postgres-formula.git ${F_DIR}/postgres
356 ( cd postgres && git checkout --quiet tags/"${POSTGRES_TAG}" )
357
358 echo "...letsencrypt"
359 test -d letsencrypt && ( cd letsencrypt && git fetch ) \
360   || git clone --quiet https://github.com/saltstack-formulas/letsencrypt-formula.git ${F_DIR}/letsencrypt
361 ( cd letsencrypt && git checkout --quiet tags/"${LETSENCRYPT_TAG}" )
362
363 echo "...logrotate"
364 test -d logrotate && ( cd logrotate && git fetch ) \
365   || git clone --quiet https://github.com/saltstack-formulas/logrotate-formula.git ${F_DIR}/logrotate
366 ( cd logrotate && git checkout --quiet tags/"${LOGROTATE_TAG}" )
367
368 echo "...arvados"
369 test -d arvados || git clone --quiet https://git.arvados.org/arvados-formula.git ${F_DIR}/arvados
370
371 # If we want to try a specific branch of the formula
372 if [ "x${BRANCH}" != "x" ]; then
373   ( cd ${F_DIR}/arvados && git checkout --quiet -t origin/"${BRANCH}" -b "${BRANCH}" )
374 elif [ "x${ARVADOS_TAG}" != "x" ]; then
375   ( cd ${F_DIR}/arvados && git checkout --quiet tags/"${ARVADOS_TAG}" -b "${ARVADOS_TAG}" )
376 fi
377
378 if [ "x${VAGRANT}" = "xyes" ]; then
379   EXTRA_STATES_DIR="/home/vagrant/${CONFIG_DIR}/states"
380   SOURCE_PILLARS_DIR="/home/vagrant/${CONFIG_DIR}/pillars"
381   SOURCE_TESTS_DIR="/home/vagrant/${TESTS_DIR}"
382 else
383   EXTRA_STATES_DIR="${SCRIPT_DIR}/${CONFIG_DIR}/states"
384   SOURCE_PILLARS_DIR="${SCRIPT_DIR}/${CONFIG_DIR}/pillars"
385   SOURCE_TESTS_DIR="${SCRIPT_DIR}/${TESTS_DIR}"
386 fi
387
388 SOURCE_STATES_DIR="${EXTRA_STATES_DIR}"
389
390 echo "Writing pillars and states"
391
392 # Replace variables (cluster,  domain, etc) in the pillars, states and tests
393 # to ease deployment for newcomers
394 if [ ! -d "${SOURCE_PILLARS_DIR}" ]; then
395   echo "${SOURCE_PILLARS_DIR} does not exist or is not a directory. Exiting."
396   exit 1
397 fi
398 for f in $(ls "${SOURCE_PILLARS_DIR}"/*); do
399   sed "s#__ANONYMOUS_USER_TOKEN__#${ANONYMOUS_USER_TOKEN}#g;
400        s#__BLOB_SIGNING_KEY__#${BLOB_SIGNING_KEY}#g;
401        s#__CONTROLLER_EXT_SSL_PORT__#${CONTROLLER_EXT_SSL_PORT}#g;
402        s#__CLUSTER__#${CLUSTER}#g;
403        s#__DOMAIN__#${DOMAIN}#g;
404        s#__HOSTNAME_EXT__#${HOSTNAME_EXT}#g;
405        s#__IP_INT__#${IP_INT}#g;
406        s#__INITIAL_USER_EMAIL__#${INITIAL_USER_EMAIL}#g;
407        s#__INITIAL_USER_PASSWORD__#${INITIAL_USER_PASSWORD}#g;
408        s#__INITIAL_USER__#${INITIAL_USER}#g;
409        s#__LE_AWS_REGION__#${LE_AWS_REGION}#g;
410        s#__LE_AWS_SECRET_ACCESS_KEY__#${LE_AWS_SECRET_ACCESS_KEY}#g;
411        s#__LE_AWS_ACCESS_KEY_ID__#${LE_AWS_ACCESS_KEY_ID}#g;
412        s#__DATABASE_PASSWORD__#${DATABASE_PASSWORD}#g;
413        s#__KEEPWEB_EXT_SSL_PORT__#${KEEPWEB_EXT_SSL_PORT}#g;
414        s#__KEEP_EXT_SSL_PORT__#${KEEP_EXT_SSL_PORT}#g;
415        s#__MANAGEMENT_TOKEN__#${MANAGEMENT_TOKEN}#g;
416        s#__RELEASE__#${RELEASE}#g;
417        s#__SYSTEM_ROOT_TOKEN__#${SYSTEM_ROOT_TOKEN}#g;
418        s#__VERSION__#${VERSION}#g;
419        s#__WEBSHELL_EXT_SSL_PORT__#${WEBSHELL_EXT_SSL_PORT}#g;
420        s#__WEBSOCKET_EXT_SSL_PORT__#${WEBSOCKET_EXT_SSL_PORT}#g;
421        s#__WORKBENCH1_EXT_SSL_PORT__#${WORKBENCH1_EXT_SSL_PORT}#g;
422        s#__WORKBENCH2_EXT_SSL_PORT__#${WORKBENCH2_EXT_SSL_PORT}#g;
423        s#__CLUSTER_INT_CIDR__#${CLUSTER_INT_CIDR}#g;
424        s#__CONTROLLER_INT_IP__#${CONTROLLER_INT_IP}#g;
425        s#__WEBSOCKET_INT_IP__#${WEBSOCKET_INT_IP}#g;
426        s#__KEEP_INT_IP__#${KEEP_INT_IP}#g;
427        s#__KEEPSTORE0_INT_IP__#${KEEPSTORE0_INT_IP}#g;
428        s#__KEEPSTORE1_INT_IP__#${KEEPSTORE1_INT_IP}#g;
429        s#__KEEPWEB_INT_IP__#${KEEPWEB_INT_IP}#g;
430        s#__WEBSHELL_INT_IP__#${WEBSHELL_INT_IP}#g;
431        s#__SHELL_INT_IP__#${SHELL_INT_IP}#g;
432        s#__WORKBENCH1_INT_IP__#${WORKBENCH1_INT_IP}#g;
433        s#__WORKBENCH2_INT_IP__#${WORKBENCH2_INT_IP}#g;
434        s#__DATABASE_INT_IP__#${DATABASE_INT_IP}#g;
435        s#__WORKBENCH_SECRET_KEY__#${WORKBENCH_SECRET_KEY}#g" \
436   "${f}" > "${P_DIR}"/$(basename "${f}")
437 done
438
439 if [ ! -d "${SOURCE_TESTS_DIR}" ]; then
440   echo "WARNING: The tests directory was not copied to \"${SOURCE_TESTS_DIR}\"."
441   if [ "x${TEST}" = "xyes" ]; then
442     echo "WARNING: Disabling tests for this installation."
443   fi
444   TEST="no"
445 else
446   mkdir -p ${T_DIR}
447   # Replace cluster and domain name in the test files
448   for f in $(ls "${SOURCE_TESTS_DIR}"/*); do
449     FILTERS="s#__CLUSTER__#${CLUSTER}#g;
450          s#__CONTROLLER_EXT_SSL_PORT__#${CONTROLLER_EXT_SSL_PORT}#g;
451          s#__DOMAIN__#${DOMAIN}#g;
452          s#__IP_INT__#${IP_INT}#g;
453          s#__INITIAL_USER_EMAIL__#${INITIAL_USER_EMAIL}#g;
454          s#__INITIAL_USER_PASSWORD__#${INITIAL_USER_PASSWORD}#g
455          s#__INITIAL_USER__#${INITIAL_USER}#g;
456          s#__DATABASE_PASSWORD__#${DATABASE_PASSWORD}#g;
457          s#__SYSTEM_ROOT_TOKEN__#${SYSTEM_ROOT_TOKEN}#g"
458     if [ "$USE_SINGLE_HOSTNAME" = "yes" ]; then
459       FILTERS="s#__CLUSTER__.__DOMAIN__#${HOSTNAME_EXT}#g;
460          $FILTERS"
461     fi
462     sed "$FILTERS" \
463       "${f}" > ${T_DIR}/$(basename "${f}")
464   done
465   chmod 755 ${T_DIR}/run-test.sh
466 fi
467
468 # Replace helper state files that differ from the formula's examples
469 if [ -d "${SOURCE_STATES_DIR}" ]; then
470   mkdir -p "${F_DIR}"/extra/extra
471
472   for f in $(ls "${SOURCE_STATES_DIR}"/*); do
473     sed "s#__ANONYMOUS_USER_TOKEN__#${ANONYMOUS_USER_TOKEN}#g;
474          s#__CLUSTER__#${CLUSTER}#g;
475          s#__BLOB_SIGNING_KEY__#${BLOB_SIGNING_KEY}#g;
476          s#__CONTROLLER_EXT_SSL_PORT__#${CONTROLLER_EXT_SSL_PORT}#g;
477          s#__DOMAIN__#${DOMAIN}#g;
478          s#__HOSTNAME_EXT__#${HOSTNAME_EXT}#g;
479          s#__IP_INT__#${IP_INT}#g;
480          s#__INITIAL_USER_EMAIL__#${INITIAL_USER_EMAIL}#g;
481          s#__INITIAL_USER_PASSWORD__#${INITIAL_USER_PASSWORD}#g;
482          s#__INITIAL_USER__#${INITIAL_USER}#g;
483          s#__DATABASE_PASSWORD__#${DATABASE_PASSWORD}#g;
484          s#__KEEPWEB_EXT_SSL_PORT__#${KEEPWEB_EXT_SSL_PORT}#g;
485          s#__KEEP_EXT_SSL_PORT__#${KEEP_EXT_SSL_PORT}#g;
486          s#__MANAGEMENT_TOKEN__#${MANAGEMENT_TOKEN}#g;
487          s#__RELEASE__#${RELEASE}#g;
488          s#__SYSTEM_ROOT_TOKEN__#${SYSTEM_ROOT_TOKEN}#g;
489          s#__VERSION__#${VERSION}#g;
490          s#__CLUSTER_INT_CIDR__#${CLUSTER_INT_CIDR}#g;
491          s#__CONTROLLER_INT_IP__#${CONTROLLER_INT_IP}#g;
492          s#__WEBSOCKET_INT_IP__#${WEBSOCKET_INT_IP}#g;
493          s#__KEEP_INT_IP__#${KEEP_INT_IP}#g;
494          s#__KEEPSTORE0_INT_IP__#${KEEPSTORE0_INT_IP}#g;
495          s#__KEEPSTORE1_INT_IP__#${KEEPSTORE1_INT_IP}#g;
496          s#__KEEPWEB_INT_IP__#${KEEPWEB_INT_IP}#g;
497          s#__WEBSHELL_INT_IP__#${WEBSHELL_INT_IP}#g;
498          s#__WORKBENCH1_INT_IP__#${WORKBENCH1_INT_IP}#g;
499          s#__WORKBENCH2_INT_IP__#${WORKBENCH2_INT_IP}#g;
500          s#__DATABASE_INT_IP__#${DATABASE_INT_IP}#g;
501          s#__WEBSHELL_EXT_SSL_PORT__#${WEBSHELL_EXT_SSL_PORT}#g;
502          s#__SHELL_INT_IP__#${SHELL_INT_IP}#g;
503          s#__WEBSOCKET_EXT_SSL_PORT__#${WEBSOCKET_EXT_SSL_PORT}#g;
504          s#__WORKBENCH1_EXT_SSL_PORT__#${WORKBENCH1_EXT_SSL_PORT}#g;
505          s#__WORKBENCH2_EXT_SSL_PORT__#${WORKBENCH2_EXT_SSL_PORT}#g;
506          s#__WORKBENCH_SECRET_KEY__#${WORKBENCH_SECRET_KEY}#g" \
507     "${f}" > "${F_DIR}/extra/extra"/$(basename "${f}")
508   done
509 fi
510
511 # Now, we build the SALT states/pillars trees
512 # As we need to separate both states and pillars in case we want specific
513 # roles, we iterate on both at the same time
514
515 # States
516 cat > ${S_DIR}/top.sls << EOFTSLS
517 base:
518   '*':
519     - locale
520 EOFTSLS
521
522 # Pillars
523 cat > ${P_DIR}/top.sls << EOFPSLS
524 base:
525   '*':
526     - locale
527     - arvados
528 EOFPSLS
529
530 # States, extra states
531 if [ -d "${F_DIR}"/extra/extra ]; then
532   SKIP_SNAKE_OIL="snakeoil_certs"
533
534   if [[ "$DEV_MODE" = "yes" || "${SSL_MODE}" == "self-signed" ]] ; then
535     # In dev mode, we create some snake oil certs that we'll
536     # use as CUSTOM_CERTS, so we don't skip the states file.
537     # Same when using self-signed certificates.
538     SKIP_SNAKE_OIL="dont_add_snakeoil_certs"
539   fi
540   for f in $(ls "${F_DIR}"/extra/extra/*.sls | egrep -v "${SKIP_SNAKE_OIL}|shell_"); do
541   echo "    - extra.$(basename ${f} | sed 's/.sls$//g')" >> ${S_DIR}/top.sls
542   done
543   # Use byo or self-signed certificates
544   if [ "${SSL_MODE}" != "lets-encrypt" ]; then
545     mkdir -p "${F_DIR}"/extra/extra/files
546   fi
547 fi
548
549 # If we want specific roles for a node, just add the desired states
550 # and its dependencies
551 if [ -z "${ROLES}" ]; then
552   # States
553   echo "    - nginx.passenger" >> ${S_DIR}/top.sls
554   if [ "${SSL_MODE}" = "lets-encrypt" ]; then
555     if [ "${USE_LETSENCRYPT_ROUTE53}" = "yes" ]; then
556       grep -q "aws_credentials" ${S_DIR}/top.sls || echo "    - extra.aws_credentials" >> ${S_DIR}/top.sls
557     fi
558     grep -q "letsencrypt"     ${S_DIR}/top.sls || echo "    - letsencrypt" >> ${S_DIR}/top.sls
559   else
560     mkdir -p /srv/salt/certs
561     if [ "${SSL_MODE}" = "bring-your-own" ]; then
562       # Copy certs to formula extra/files
563       cp -rv ${CUSTOM_CERTS_DIR}/* /srv/salt/certs/
564       # We add the custom_certs state
565       grep -q "custom_certs"    ${S_DIR}/top.sls || echo "    - extra.custom_certs" >> ${S_DIR}/top.sls
566     fi
567     # In self-signed mode, the certificate files will be created and put in the
568     # destination directory by the snakeoil_certs.sls state file
569   fi
570
571   echo "    - postgres" >> ${S_DIR}/top.sls
572   echo "    - logrotate" >> ${S_DIR}/top.sls
573   echo "    - docker.software" >> ${S_DIR}/top.sls
574   echo "    - arvados" >> ${S_DIR}/top.sls
575   echo "    - extra.shell_sudo_passwordless" >> ${S_DIR}/top.sls
576   echo "    - extra.shell_cron_add_login_sync" >> ${S_DIR}/top.sls
577   echo "    - extra.passenger_rvm" >> ${S_DIR}/top.sls
578
579   # Pillars
580   echo "    - docker" >> ${P_DIR}/top.sls
581   echo "    - nginx_api_configuration" >> ${P_DIR}/top.sls
582   echo "    - logrotate_api" >> ${P_DIR}/top.sls
583   echo "    - nginx_controller_configuration" >> ${P_DIR}/top.sls
584   echo "    - nginx_keepproxy_configuration" >> ${P_DIR}/top.sls
585   echo "    - nginx_keepweb_configuration" >> ${P_DIR}/top.sls
586   echo "    - nginx_passenger" >> ${P_DIR}/top.sls
587   echo "    - nginx_websocket_configuration" >> ${P_DIR}/top.sls
588   echo "    - nginx_webshell_configuration" >> ${P_DIR}/top.sls
589   echo "    - nginx_workbench2_configuration" >> ${P_DIR}/top.sls
590   echo "    - nginx_workbench_configuration" >> ${P_DIR}/top.sls
591   echo "    - logrotate_wb1" >> ${P_DIR}/top.sls
592   echo "    - postgresql" >> ${P_DIR}/top.sls
593
594   # We need to tweak the Nginx's pillar depending whether we want plan nginx or nginx+passenger
595   NGINX_INSTALL_SOURCE="install_from_phusionpassenger"
596   sed -i "s/__NGINX_INSTALL_SOURCE__/${NGINX_INSTALL_SOURCE}/g" ${P_DIR}/nginx_passenger.sls
597
598   if [ "${SSL_MODE}" = "lets-encrypt" ]; then
599     if [ "${USE_LETSENCRYPT_ROUTE53}" = "yes" ]; then
600       grep -q "aws_credentials" ${P_DIR}/top.sls || echo "    - aws_credentials" >> ${P_DIR}/top.sls
601     fi
602     grep -q "letsencrypt" ${P_DIR}/top.sls || echo "    - letsencrypt" >> ${P_DIR}/top.sls
603
604     hosts=("controller" "websocket" "workbench" "workbench2" "webshell" "keepproxy")
605     if [ ${USE_SINGLE_HOSTNAME} = "no" ]; then
606       hosts+=("download" "collections")
607     else
608       hosts+=("keepweb")
609     fi
610
611     for c in "${hosts[@]}"; do
612       # Are we in a single-host-single-hostname env?
613       if [ "${USE_SINGLE_HOSTNAME}" = "yes" ]; then
614         # Are we in a single-host-single-hostname env?
615         CERT_NAME=${HOSTNAME_EXT}
616       else
617         # We are in a multiple-hostnames env
618         CERT_NAME=${c}.${CLUSTER}.${DOMAIN}
619       fi
620
621       # As the pillar differs whether we use LE or custom certs, we need to do a final edition on them
622       sed -i "s/__CERT_REQUIRES__/cmd: create-initial-cert-${CERT_NAME}*/g;
623               s#__CERT_PEM__#/etc/letsencrypt/live/${CERT_NAME}/fullchain.pem#g;
624               s#__CERT_KEY__#/etc/letsencrypt/live/${CERT_NAME}/privkey.pem#g" \
625       ${P_DIR}/nginx_${c}_configuration.sls
626     done
627   else
628     # Use custom certs (either dev mode or prod)
629     grep -q "extra_custom_certs" ${P_DIR}/top.sls || echo "    - extra_custom_certs" >> ${P_DIR}/top.sls
630     # And add the certs in the custom_certs pillar
631     echo "extra_custom_certs_dir: /srv/salt/certs" > ${P_DIR}/extra_custom_certs.sls
632     echo "extra_custom_certs:" >> ${P_DIR}/extra_custom_certs.sls
633
634     for c in controller websocket workbench workbench2 webshell keepweb keepproxy shell; do
635       # Are we in a single-host-single-hostname env?
636       if [ "${USE_SINGLE_HOSTNAME}" = "yes" ]; then
637         # Are we in a single-host-single-hostname env?
638         CERT_NAME=${HOSTNAME_EXT}
639       else
640         # We are in a multiple-hostnames env
641         CERT_NAME=${c}
642       fi
643
644       if [[ "$SSL_MODE" == "bring-your-own" ]]; then
645         copy_custom_cert ${CUSTOM_CERTS_DIR} ${CERT_NAME}
646       fi
647
648       grep -q ${CERT_NAME} ${P_DIR}/extra_custom_certs.sls || echo "  - ${CERT_NAME}" >> ${P_DIR}/extra_custom_certs.sls
649
650       # As the pillar differs whether we use LE or custom certs, we need to do a final edition on them
651       sed -i "s/__CERT_REQUIRES__/file: extra_custom_certs_file_copy_arvados-${CERT_NAME}.pem/g;
652               s#__CERT_PEM__#/etc/nginx/ssl/arvados-${CERT_NAME}.pem#g;
653               s#__CERT_KEY__#/etc/nginx/ssl/arvados-${CERT_NAME}.key#g" \
654       ${P_DIR}/nginx_${c}_configuration.sls
655     done
656   fi
657 else
658   # If we add individual roles, make sure we add the repo first
659   echo "    - arvados.repo" >> ${S_DIR}/top.sls
660   # We add the extra_custom_certs state
661   grep -q "extra.custom_certs"    ${S_DIR}/top.sls || echo "    - extra.custom_certs" >> ${S_DIR}/top.sls
662
663   # And we add the basic part for the certs pillar
664   if [ "${SSL_MODE}" != "lets-encrypt" ]; then
665     # And add the certs in the custom_certs pillar
666     echo "extra_custom_certs_dir: /srv/salt/certs" > ${P_DIR}/extra_custom_certs.sls
667     echo "extra_custom_certs:" >> ${P_DIR}/extra_custom_certs.sls
668     grep -q "extra_custom_certs" ${P_DIR}/top.sls || echo "    - extra_custom_certs" >> ${P_DIR}/top.sls
669   fi
670
671   for R in ${ROLES}; do
672     case "${R}" in
673       "database")
674         # States
675         echo "    - postgres" >> ${S_DIR}/top.sls
676         # Pillars
677         echo '    - postgresql' >> ${P_DIR}/top.sls
678       ;;
679       "api")
680         # States
681         grep -q "    - logrotate" ${S_DIR}/top.sls || echo "    - logrotate" >> ${S_DIR}/top.sls
682         if grep -q "    - nginx.*$" ${S_DIR}/top.sls; then
683           sed -i s/"^    - nginx.*$"/"    - nginx.passenger"/g ${S_DIR}/top.sls
684         else
685           echo "    - nginx.passenger" >> ${S_DIR}/top.sls
686         fi
687         echo "    - extra.passenger_rvm" >> ${S_DIR}/top.sls
688         ### If we don't install and run LE before arvados-api-server, it fails and breaks everything
689         ### after it. So we add this here as we are, after all, sharing the host for api and controller
690         if [ "${SSL_MODE}" = "lets-encrypt" ]; then
691           if [ "${USE_LETSENCRYPT_ROUTE53}" = "yes" ]; then
692             grep -q "aws_credentials" ${S_DIR}/top.sls || echo "    - aws_credentials" >> ${S_DIR}/top.sls
693           fi
694           grep -q "letsencrypt" ${S_DIR}/top.sls || echo "    - letsencrypt" >> ${S_DIR}/top.sls
695         else
696           # Use custom certs
697           if [ "${SSL_MODE}" = "bring-your-own" ]; then
698             copy_custom_cert ${CUSTOM_CERTS_DIR} controller
699           fi
700           grep -q controller ${P_DIR}/extra_custom_certs.sls || echo "  - controller" >> ${P_DIR}/extra_custom_certs.sls
701         fi
702         grep -q "arvados.${R}" ${S_DIR}/top.sls    || echo "    - arvados.${R}" >> ${S_DIR}/top.sls
703         # Pillars
704         grep -q "logrotate_api" ${P_DIR}/top.sls            || echo "    - logrotate_api" >> ${P_DIR}/top.sls
705         grep -q "aws_credentials" ${P_DIR}/top.sls          || echo "    - aws_credentials" >> ${P_DIR}/top.sls
706         grep -q "postgresql" ${P_DIR}/top.sls               || echo "    - postgresql" >> ${P_DIR}/top.sls
707         grep -q "nginx_passenger" ${P_DIR}/top.sls          || echo "    - nginx_passenger" >> ${P_DIR}/top.sls
708         grep -q "nginx_${R}_configuration" ${P_DIR}/top.sls || echo "    - nginx_${R}_configuration" >> ${P_DIR}/top.sls
709
710         # We need to tweak the Nginx's pillar depending whether we want plain nginx or nginx+passenger
711         NGINX_INSTALL_SOURCE="install_from_phusionpassenger"
712         sed -i "s/__NGINX_INSTALL_SOURCE__/${NGINX_INSTALL_SOURCE}/g" ${P_DIR}/nginx_passenger.sls
713       ;;
714       "controller" | "websocket" | "workbench" | "workbench2" | "webshell" | "keepweb" | "keepproxy")
715         # States
716         if [ "${R}" = "workbench" ]; then
717           grep -q "    - logrotate" ${S_DIR}/top.sls || echo "    - logrotate" >> ${S_DIR}/top.sls
718           NGINX_INSTALL_SOURCE="install_from_phusionpassenger"
719           if grep -q "    - nginx$" ${S_DIR}/top.sls; then
720             sed -i s/"^    - nginx.*$"/"    - nginx.passenger"/g ${S_DIR}/top.sls
721           else
722             echo "    - nginx.passenger" >> ${S_DIR}/top.sls
723           fi
724         else
725           grep -q "nginx" ${S_DIR}/top.sls || echo "    - nginx" >> ${S_DIR}/top.sls
726         fi
727         if [ "${SSL_MODE}" = "lets-encrypt" ]; then
728           if [ "x${USE_LETSENCRYPT_ROUTE53}" = "xyes" ]; then
729             grep -q "aws_credentials" ${S_DIR}/top.sls || echo "    - aws_credentials" >> ${S_DIR}/top.sls
730           fi
731           grep -q "letsencrypt"     ${S_DIR}/top.sls || echo "    - letsencrypt" >> ${S_DIR}/top.sls
732         else
733           # Use custom certs, special case for keepweb
734           if [ ${R} = "keepweb" ]; then
735             if [ "${SSL_MODE}" = "bring-your-own" ]; then
736               copy_custom_cert ${CUSTOM_CERTS_DIR} download
737               copy_custom_cert ${CUSTOM_CERTS_DIR} collections
738             fi
739           else
740             if [ "${SSL_MODE}" = "bring-your-own" ]; then
741               copy_custom_cert ${CUSTOM_CERTS_DIR} ${R}
742             fi
743           fi
744         fi
745         # webshell role is just a nginx vhost, so it has no state
746         if [ "${R}" != "webshell" ]; then
747           grep -q "arvados.${R}" ${S_DIR}/top.sls || echo "    - arvados.${R}" >> ${S_DIR}/top.sls
748         fi
749         # Pillars
750         if [ "${R}" = "workbench" ]; then
751           grep -q "logrotate_wb1" ${P_DIR}/top.sls || echo "    - logrotate_wb1" >> ${P_DIR}/top.sls
752         fi
753         grep -q "nginx_passenger" ${P_DIR}/top.sls          || echo "    - nginx_passenger" >> ${P_DIR}/top.sls
754         grep -q "nginx_${R}_configuration" ${P_DIR}/top.sls || echo "    - nginx_${R}_configuration" >> ${P_DIR}/top.sls
755         # Special case for keepweb
756         if [ ${R} = "keepweb" ]; then
757           grep -q "nginx_download_configuration" ${P_DIR}/top.sls || echo "    - nginx_download_configuration" >> ${P_DIR}/top.sls
758           grep -q "nginx_collections_configuration" ${P_DIR}/top.sls || echo "    - nginx_collections_configuration" >> ${P_DIR}/top.sls
759         fi
760
761         if [ "${SSL_MODE}" = "lets-encrypt" ]; then
762           if [ "${USE_LETSENCRYPT_ROUTE53}" = "yes" ]; then
763             grep -q "aws_credentials" ${P_DIR}/top.sls || echo "    - aws_credentials" >> ${P_DIR}/top.sls
764           fi
765           grep -q "letsencrypt"     ${P_DIR}/top.sls || echo "    - letsencrypt" >> ${P_DIR}/top.sls
766           grep -q "letsencrypt_${R}_configuration" ${P_DIR}/top.sls || echo "    - letsencrypt_${R}_configuration" >> ${P_DIR}/top.sls
767
768           # As the pillar differ whether we use LE or custom certs, we need to do a final edition on them
769           # Special case for keepweb
770           if [ ${R} = "keepweb" ]; then
771             for kwsub in download collections; do
772               sed -i "s/__CERT_REQUIRES__/cmd: create-initial-cert-${kwsub}.${CLUSTER}.${DOMAIN}*/g;
773                       s#__CERT_PEM__#/etc/letsencrypt/live/${kwsub}.${CLUSTER}.${DOMAIN}/fullchain.pem#g;
774                       s#__CERT_KEY__#/etc/letsencrypt/live/${kwsub}.${CLUSTER}.${DOMAIN}/privkey.pem#g" \
775               ${P_DIR}/nginx_${kwsub}_configuration.sls
776             done
777           else
778             sed -i "s/__CERT_REQUIRES__/cmd: create-initial-cert-${R}.${CLUSTER}.${DOMAIN}*/g;
779                     s#__CERT_PEM__#/etc/letsencrypt/live/${R}.${CLUSTER}.${DOMAIN}/fullchain.pem#g;
780                     s#__CERT_KEY__#/etc/letsencrypt/live/${R}.${CLUSTER}.${DOMAIN}/privkey.pem#g" \
781             ${P_DIR}/nginx_${R}_configuration.sls
782           fi
783         else
784           # As the pillar differ whether we use LE or custom certs, we need to do a final edition on them
785           # Special case for keepweb
786           if [ ${R} = "keepweb" ]; then
787             for kwsub in download collections; do
788               sed -i "s/__CERT_REQUIRES__/file: extra_custom_certs_file_copy_arvados-${kwsub}.pem/g;
789                       s#__CERT_PEM__#/etc/nginx/ssl/arvados-${kwsub}.pem#g;
790                       s#__CERT_KEY__#/etc/nginx/ssl/arvados-${kwsub}.key#g" \
791               ${P_DIR}/nginx_${kwsub}_configuration.sls
792               grep -q ${kwsub} ${P_DIR}/extra_custom_certs.sls || echo "  - ${kwsub}" >> ${P_DIR}/extra_custom_certs.sls
793             done
794           else
795             sed -i "s/__CERT_REQUIRES__/file: extra_custom_certs_file_copy_arvados-${R}.pem/g;
796                     s#__CERT_PEM__#/etc/nginx/ssl/arvados-${R}.pem#g;
797                     s#__CERT_KEY__#/etc/nginx/ssl/arvados-${R}.key#g" \
798             ${P_DIR}/nginx_${R}_configuration.sls
799             grep -q ${R} ${P_DIR}/extra_custom_certs.sls || echo "  - ${R}" >> ${P_DIR}/extra_custom_certs.sls
800           fi
801         fi
802         # We need to tweak the Nginx's pillar depending whether we want plain nginx or nginx+passenger
803         sed -i "s/__NGINX_INSTALL_SOURCE__/${NGINX_INSTALL_SOURCE}/g" ${P_DIR}/nginx_passenger.sls
804       ;;
805       "shell")
806         # States
807         echo "    - extra.shell_sudo_passwordless" >> ${S_DIR}/top.sls
808         echo "    - extra.shell_cron_add_login_sync" >> ${S_DIR}/top.sls
809         grep -q "docker" ${S_DIR}/top.sls       || echo "    - docker.software" >> ${S_DIR}/top.sls
810         grep -q "arvados.${R}" ${S_DIR}/top.sls || echo "    - arvados.${R}" >> ${S_DIR}/top.sls
811         # Pillars
812         grep -q "docker" ${P_DIR}/top.sls       || echo "    - docker" >> ${P_DIR}/top.sls
813       ;;
814       "dispatcher" | "keepbalance" | "keepstore")
815         # States
816         grep -q "arvados.${R}" ${S_DIR}/top.sls || echo "    - arvados.${R}" >> ${S_DIR}/top.sls
817         # Pillars
818         # ATM, no specific pillar needed
819       ;;
820       *)
821         echo "Unknown role ${R}"
822         exit 1
823       ;;
824     esac
825   done
826 fi
827
828 if [ "${DUMP_CONFIG}" = "yes" ]; then
829   # We won't run the rest of the script because we're just dumping the config
830   exit 0
831 fi
832
833 # Now run the install
834 salt-call --local state.apply -l ${LOG_LEVEL}
835
836 # Finally, make sure that /etc/hosts is not overwritten on reboot
837 if [ -d /etc/cloud/cloud.cfg.d ]; then
838   # TODO: will this work on CentOS?
839   sed -i 's/^manage_etc_hosts: true/#manage_etc_hosts: true/g' /etc/cloud/cloud.cfg.d/*
840 fi
841
842 # Leave a copy of the Arvados CA so the user can copy it where it's required
843 if [ "$DEV_MODE" = "yes" ]; then
844   echo "Copying the Arvados CA certificate to the installer dir, so you can import it"
845   # If running in a vagrant VM, also add default user to docker group
846   if [ "x${VAGRANT}" = "xyes" ]; then
847     cp /etc/ssl/certs/arvados-snakeoil-ca.pem /vagrant/${CLUSTER}.${DOMAIN}-arvados-snakeoil-ca.pem
848
849     echo "Adding the vagrant user to the docker group"
850     usermod -a -G docker vagrant
851   else
852     cp /etc/ssl/certs/arvados-snakeoil-ca.pem ${SCRIPT_DIR}/${CLUSTER}.${DOMAIN}-arvados-snakeoil-ca.pem
853   fi
854 fi
855
856 # Test that the installation finished correctly
857 if [ "x${TEST}" = "xyes" ]; then
858   cd ${T_DIR}
859   # If we use RVM, we need to run this with it, or most ruby commands will fail
860   RVM_EXEC=""
861   if [ -x /usr/local/rvm/bin/rvm-exec ]; then
862     RVM_EXEC="/usr/local/rvm/bin/rvm-exec"
863   fi
864   ${RVM_EXEC} ./run-test.sh
865 fi