On Jul 15, 2013, at 6:29 PM, Seth Sims wrote: > After some work i've created an Upstart script which can manage a load balanced galaxy configuration as described in the wiki. I thought that I would put it on this list for other people to use. The script parses universe_wsgi.ini just like run.sh and spawns all of the servers it finds. It comes in two pieces galaxy.conf and galaxy-worker.conf. Once you place them both in /etc/init and make the proper edits for the environment a server can be started with "sudo start galaxy". The configuration starts the server at boot time and puts all of the instances under the management of upstart which deals with pids, logging to syslog and respawning if an instance crashes. > I have just gotten this working reasonably well but have done basically no testing so there are bugs to be found. Any comments are welcome if anyone knows a better way to do something here. > > - Seth Hi Seth, Thanks for submitting these. I was about to commit them to the contrib/ directory along with the rest of the start scripts, but I was wondering if you could avoid running the check/fetch scripts as root by just using `su -c`? --nate > > *galaxy.conf* > ---------------------------- > author "Seth Sims <seth.sims(a)gmail.com&gt;&quot; > version "0.0.1 test" > description "galaxy master process. Fetches eggs and spawns all of the servers it finds configured" > > start on started network-services > > # make sure that any eggs we download are at least owned by the galaxy group. > # we cannot use setuid in this script because only root can issue the "start galaxy-worker" > # command. But this way the galaxy instances should still be able to use their eggs. > setgid galaxy > > # put galaxy root directory here > chdir /srv/galaxy-dist/ > > pre-start script > date > echo "checking python version" > python ./scripts/check_python.py > [ $? -ne 0 ] && exit 1 > > echo "pre-fetching tossing out expired eggs" > python ./scripts/check_eggs.py -q > if [ $? -eq 0 ]; then > echo "Some eggs are out of date, attempting to fetch..." > python ./scripts/fetch_eggs.py > if [ $? -eq 0 ]; then > echo "Fetch Successful." > else > echo "Fetch failed." > fi > fi > > echo "starting servers" > SERVERS=`sed -n 's/^\[server:\(.*\)\]/\1/ p' universe_wsgi.ini | xargs echo` > for SERVER in ${SERVERS} ; do > echo "starting server ${SERVER}" > start galaxy-worker SERVER_NAME=${SERVER} > done > end script > > post-stop script > SERVERS=`sed -n 's/^\[server:\(.*\)\]/\1/ p' universe_wsgi.ini | xargs echo` > date > echo "stopping galaxy servers". > for SERVER in ${SERVERS} ; do > echo "stopping ${SERVER}" > stop galaxy-worker SERVER_NAME=${SERVER} > done > end script > --------------------------- > *galaxy-worker* > author "Seth Sims <seth.sims(a)gmail.com&gt;&quot; > version "0.0.1 test" > description "Starts a galaxy server instance. This is run from the galaxy.conf file" > > instance $SERVER_NAME > > #make sure we are running as the galaxy user > setuid galaxy > setgid galaxy > > #put the galaxy root directory here > chdir /srv/galaxy-dist/ > > #having multiple instances of galaxy using the same egg directory was causing a race > #condition that was stopping the instances from starting correctly. So give each instance > #its own directory under /tmp > env PYTHON_EGG_CACHE=/tmp/${SERVER_NAME}_egg/ > > respawn > > script > exec python ./scripts/paster.py serve universe_wsgi.ini --server-name=${SERVER_NAME} > end script > ___________________________________________________________ > Please keep all replies on the list by using "reply all" > in your mail client. To manage your subscriptions to this > and other Galaxy lists, please use the interface at: > http://lists.bx.psu.edu/ > > To search Galaxy mailing lists use the unified search at: > http://galaxyproject.org/search/mailinglists/

On Aug 9, 2013, at 11:53 AM, Seth Sims wrote: > Dear Nate, > > Adding "su - galaxy" as the first line of the pre-start script seems to work reasonably well. Also it looks like the line that sets the egg cache is not working properly. My egg cache ends up being "/tmp/${SERVER_NAME}_egg/" but things still seem to be working so I've changed that part to use one directory in /tmp/ for all instances. Hi Seth, Is your Galaxy user's home directory /srv/galaxy-dist? Otherwise, `su - galaxy` would change the working directory to that user's home directory and the rest of the script would fail. I was thinking it might work to just use `su -c` for individual commands, e.g.: pre-start script echo "checking python version" su - galaxy -c "cd /srv/galaxy-dist ; python ./scripts/check_python.py" [ $? -ne 0 ] && exit 1 echo "pre-fetching tossing out expired eggs" su - galaxy -c "cd /srv/galaxy-dist ; python ./scripts/check_eggs.py -q" if [ $? -eq 0 ]; then echo "Some eggs are out of date, attempting to fetch..." su - galaxy -c "cd /srv/galaxy-dist ; python ./scripts/fetch_eggs.py" if [ $? -eq 0 ]; then echo "Fetch Successful." else echo "Fetch failed." fi fi echo "starting servers" SERVERS=`sed -n 's/^\[server:\(.*\)\]/\1/ p' universe_wsgi.ini | xargs echo` for SERVER in ${SERVERS} ; do echo "starting server ${SERVER}" start galaxy-worker SERVER_NAME=${SERVER} done end script > > Sincerely, > Seth Sims > > *galaxy.conf* > ---------------------------- > author "Seth Sims <seth.sims(a)gmail.com&gt;&quot; > version "0.0.2" > description "galaxy master process. Fetches eggs and spawns all of the servers it finds configured" > > start on started network-services > > # put galaxy root directory here > chdir /srv/galaxy-dist/ > > pre-start script > su - galaxy > date > echo "checking python version" > python ./scripts/check_python.py > [ $? -ne 0 ] && exit 1 > > echo "pre-fetching tossing out expired eggs" > python ./scripts/check_eggs.py -q > if [ $? -eq 0 ]; then > echo "Some eggs are out of date, attempting to fetch..." > python ./scripts/fetch_eggs.py > if [ $? -eq 0 ]; then > echo "Fetch Successful." > else > echo "Fetch failed." > fi > fi > > echo "starting servers" > SERVERS=`sed -n 's/^\[server:\(.*\)\]/\1/ p' universe_wsgi.ini | xargs echo` > for SERVER in ${SERVERS} ; do > echo "starting server ${SERVER}" > start galaxy-worker SERVER_NAME=${SERVER} > done > end script > > post-stop script > SERVERS=`sed -n 's/^\[server:\(.*\)\]/\1/ p' universe_wsgi.ini | xargs echo` > date > echo "stopping galaxy servers". > for SERVER in ${SERVERS} ; do > echo "stopping ${SERVER}" > stop galaxy-worker SERVER_NAME=${SERVER} > done > end script > --------------------------- > *galaxy-worker* > author "Seth Sims <seth.sims(a)gmail.com&gt;&quot; > version "0.0.2" > description "Starts a galaxy server instance. This is run from the galaxy.conf file" > > instance $SERVER_NAME > > #make sure we are running as the galaxy user > setuid galaxy > setgid galaxy > > #put the galaxy root directory here > chdir /srv/galaxy-dist/ > > #system users don't have /home/ directories; so point the egg unpack to a directory in /tmp/ > env PYTHON_EGG_CACHE=/tmp/galaxy_eggs/ > > respawn > > script > exec python ./scripts/paster.py serve universe_wsgi.ini --server-name=${SERVER_NAME} > end script > > > On Thu, Aug 8, 2013 at 1:52 PM, Nate Coraor <nate(a)bx.psu.edu&gt; wrote: > On Jul 15, 2013, at 6:29 PM, Seth Sims wrote: > > > After some work i've created an Upstart script which can manage a load balanced galaxy configuration as described in the wiki. I thought that I would put it on this list for other people to use. The script parses universe_wsgi.ini just like run.sh and spawns all of the servers it finds. It comes in two pieces galaxy.conf and galaxy-worker.conf. Once you place them both in /etc/init and make the proper edits for the environment a server can be started with "sudo start galaxy". The configuration starts the server at boot time and puts all of the instances under the management of upstart which deals with pids, logging to syslog and respawning if an instance crashes. > > I have just gotten this working reasonably well but have done basically no testing so there are bugs to be found. Any comments are welcome if anyone knows a better way to do something here. > > > > - Seth > > Hi Seth, > > Thanks for submitting these. I was about to commit them to the contrib/ directory along with the rest of the start scripts, but I was wondering if you could avoid running the check/fetch scripts as root by just using `su -c`? > > --nate > > > > > *galaxy.conf* > > ---------------------------- > > author "Seth Sims <seth.sims(a)gmail.com&gt;&quot; > > version "0.0.1 test" > > description "galaxy master process. Fetches eggs and spawns all of the servers it finds configured" > > > > start on started network-services > > > > # make sure that any eggs we download are at least owned by the galaxy group. > > # we cannot use setuid in this script because only root can issue the "start galaxy-worker" > > # command. But this way the galaxy instances should still be able to use their eggs. > > setgid galaxy > > > > # put galaxy root directory here > > chdir /srv/galaxy-dist/ > > > > pre-start script > > date > > echo "checking python version" > > python ./scripts/check_python.py > > [ $? -ne 0 ] && exit 1 > > > > echo "pre-fetching tossing out expired eggs" > > python ./scripts/check_eggs.py -q > > if [ $? -eq 0 ]; then > > echo "Some eggs are out of date, attempting to fetch..." > > python ./scripts/fetch_eggs.py > > if [ $? -eq 0 ]; then > > echo "Fetch Successful." > > else > > echo "Fetch failed." > > fi > > fi > > > > echo "starting servers" > > SERVERS=`sed -n 's/^\[server:\(.*\)\]/\1/ p' universe_wsgi.ini | xargs echo` > > for SERVER in ${SERVERS} ; do > > echo "starting server ${SERVER}" > > start galaxy-worker SERVER_NAME=${SERVER} > > done > > end script > > > > post-stop script > > SERVERS=`sed -n 's/^\[server:\(.*\)\]/\1/ p' universe_wsgi.ini | xargs echo` > > date > > echo "stopping galaxy servers". > > for SERVER in ${SERVERS} ; do > > echo "stopping ${SERVER}" > > stop galaxy-worker SERVER_NAME=${SERVER} > > done > > end script > > --------------------------- > > *galaxy-worker* > > author "Seth Sims <seth.sims(a)gmail.com&gt;&quot; > > version "0.0.1 test" > > description "Starts a galaxy server instance. This is run from the galaxy.conf file" > > > > instance $SERVER_NAME > > > > #make sure we are running as the galaxy user > > setuid galaxy > > setgid galaxy > > > > #put the galaxy root directory here > > chdir /srv/galaxy-dist/ > > > > #having multiple instances of galaxy using the same egg directory was causing a race > > #condition that was stopping the instances from starting correctly. So give each instance > > #its own directory under /tmp > > env PYTHON_EGG_CACHE=/tmp/${SERVER_NAME}_egg/ > > > > respawn > > > > script > > exec python ./scripts/paster.py serve universe_wsgi.ini --server-name=${SERVER_NAME} > > end script > > ___________________________________________________________ > > Please keep all replies on the list by using "reply all" > > in your mail client. To manage your subscriptions to this > > and other Galaxy lists, please use the interface at: > > http://lists.bx.psu.edu/ > > > > To search Galaxy mailing lists use the unified search at: > > http://galaxyproject.org/search/mailinglists/ > >