Difference between revisions of "Administration:Installing WormMine"

From WormBaseWiki
Jump to navigationJump to search
Line 587: Line 587:
  
 
=== Launch webapp ===
 
=== Launch webapp ===
 +
* Note: Catalina must be running before you deploy the webapp ([[Administration:Installing_WormMine#About_Catalina |about catalina]])
 
* Navigate to intermine/wormmine/webapp
 
* Navigate to intermine/wormmine/webapp
 
* Launch webapp:
 
* Launch webapp:

Revision as of 18:36, 31 October 2013

How to set up a development instance of WormMine

Contents

Quick Overview

Data flow diagram
Branch diagram
  • InterMine machines:
    • production: 206.108.125.166
    • development: 206.108.125.174
  • Important directories:
    • database backup: /nfs/wormbase/wormmine/database_dumps
    • in intermine directory:
      • datadir: holds data files
      • build_config: config files for build
      • redeployment: mine update build

Login information is available in a mine's ~/.intermine/wormmine.properties file.

Please compile a list of the files whose contents need to be updated when incrementing the version number. For example, "website-intermine/acedb-dev/intermine/wormmine/project.xml" needs to be updated because it contains the string "WS239".

Requirements

Hardware

Linux

  • 8 cores
  • 24GB RAM
  • ~ 1TB storage

Software

Necessary software and versions:

Software Minimum Version Purpose
Git 1.7 check out and update source code
Java SDK 6.0 build and use InterMine
Ant 1.8 invokes the InterMine build
Tomcat 6.0.29 website
PostgreSQL 8.3 database
Perl 5.8.8 run build scripts

Installation / configuration

Dependencies

Git

Install the command line tool:

$ sudo apt-get install git-core

Configure your user and email:

$ git config --global user.name "Name Surname"
$ git config --global user.email "your.email@gmail.com"

Java

Download here. Since InterMine can be memory intensive, it's helpful to pass environment variables to ant through the ANT_OPTS variable.

TODO: figure out the real reason why these parameters should be used. They have little to do with memory usage, but appear to optimize for throughput.

$ export ANT_OPTS="-server -XX:MaxPermSize=256M -Xmx1700m -XX:+UseParallelGC
-Xms1700m -XX:SoftRefLRUPolicyMSPerMB=1 -XX:MaxHeapFreeRatio=99"

Ant

Refer to ant's manual for installation instructions.

Tomcat

Refer to Tomcat InterMine installation

Post-Installation Environment

export CATALINA_HOME=/YOURPATH/apache-tomcat-6.0.36

Configuring an Alternative HTTP Port

vim apache-tomcat-6.0.36/conf/server.xml

Replace the port number in this context:

<Service name="Catalina">
    <Connector port="YOURPORT" protocol="HTTP/1.1"
               connectionTimeout="20000"
               redirectPort="8443"
               URIEncoding="UTF-8" />
...

For user-specific port numbers see the Google Doc "Developer Resources at WormBase".

Starting Tomcat

$CATALINA_HOME/bin/startup.sh

Stopping Tomcat

$CATALINA_HOME/bin/shutdown.sh

PostgreSQL

Refer to InterMine PostgreSQL installation guide

Perl

Refer to InterMine Perl installation guide

Download and Install WormMine

Navigate into the folder you want to install WormMine

Download code from Git:

~]$ git clone https://github.com/WormBase/website-intermine.git
Cloning into website-intermine...
remote: Counting objects: 2866, done.
remote: Compressing objects: 100% (1054/1054), done.
remote: Total 2866 (delta 1728), reused 2837 (delta 1699)
Receiving objects: 100% (2866/2866), 25.28 MiB | 4.65 MiB/s, done.
Resolving deltas: 100% (1728/1728), done.

This downloads the entire intermine project repository. The mine itself is a submodule of this.

Initialize submodule

~]$ cd website-intermine
website-intermine]$ git submodule update --init
Submodule 'acedb-dev/intermine' (git@github.com:WormBase/intermine.git) registered for path 'acedb-dev/intermine'
Cloning into acedb-dev/intermine...
Warning: Permanently added the RSA host key for IP address '192.30.252.131' to the list of known hosts.
remote: Counting objects: 379065, done.
remote: Compressing objects: 100% (78659/78659), done.
remote: Total 379065 (delta 233912), reused 377610 (delta 232792)
Receiving objects: 100% (379065/379065), 685.14 MiB | 6.47 MiB/s, done.
Resolving deltas: 100% (233912/233912), done. 
Submodule path 'acedb-dev/intermine': checked out 'd640534eda614d60558c6561da6fb9311d6ad893'

This populates the intermine directory at website-intermine/acedb-dev/intermine It needs to be set to the proper branch.

Navigate to the mine:

website-intermine]$ cd acedb-dev/intermine/

Working with the Development Branch

Get the "website-intermine" repo as before, but do not init the submodule. Instead proceed as follows:

git clone https://github.com/WormBase/website-intermine.git
git clone https://github.com/WormBase/intermine

# Create link to the InterMine sources within the web-site context. Replaces sub-module.
cd website-intermine/acedb-dev
rmdir intermine
ln -s ../../intermine .
cd ../../intermine

# Get development branch.
git remote add unmerged https://github.com/WormBase/intermine/tree/unmerged
git checkout unmerged

Create properties file

  • Create ~/.intermine directory.
  • Copy the sample properties file as ~/.intermine/wormmine.properties
  • Fill in placeholders as follows:
    • <POSTGRES USER PASSWORD>: postgres password for intermine user
    • <TOMCAT USER PASSWORD>: tomcat password for intermine user
    • <SERVER PUBLIC BASE URL>: Base url of your web server, including port. Sample: http://123.456.789.123:8080
    • <CREATE WM ADMIN USERNAME/PASSWORD> create the primary admin account
    • <EMAIL ADDRESS TO SEND HELP EMAILS FROM>: your server should be configured to send emails from this address. This will send users password reset emails and the like.
    • <HELP REQUESTS ARE SENT HERE>: can be same address as above, this is where input from the InterMine help form gets sent.

Get production database

Build a new production database

Custom WormBase loading processes

Protein Fasta

  • custom loader
    WormBaseProteinFastaLoaderTask.java
  • puts the first ID of the title row into both Protein.primaryIdentifier and Protein.primaryAccession, and the second ID into protein.secondaryIdentifier

GFF3

  • custom loader
    WormbaseGff3CoreGFF3RecordHandler.java
  • when processing transcript record it creates a reference to it's parent gene
  • when processing coding sequence record it creates a reference to it's parent transcript

XML

  • custom loader
    WormbaseAcedbConverter.java
  • can load any XML, not ace specific
  • loads intermine class fields as mapped by mapping file

Mapping file format:

# this is a comment, comment lines are ignored

# regular annotation, this field will be filled in 
# with the value of the evaluated xpath
primaryIdentifier = /Variation/text()[1]

# returns true if xpath returns any nodes at all, 
# useful for data contained in ace tags themselves
if.naturalVariant = /XPATH/...

# type casting allowed, this example will add the value as a
# Phenotype record
(Phenotype)parents = /XPATH/... 

The build requires:

  • Fasta
  • GFF3
  • GO ontology
  • GO association
  • Ace XML files

All of these are retrieved from the FTP site, except for the Ace XML files.

Generate data files

These must be created on a machine with a tace instance, and stored in an accessible location.

Generate Ace XML files

Both the blanket dump and manual queries must be run and saved to XML to represent all covered types

Blanket dump

On machine with Ace instance.

  • Download the website-intermine repository as described above
  • Navigate to acedb-dev/acedb
website-intermine]$ cd acedb-dev/acedb/

imdump.sh is a shell script which generates XML files for each species in model, into the supplied destination directory. The intermine machine must have access to the directory.

Run it with: any folder can be used as the Ace XML dump location

>  ./imdump.sh <ACE XML DUMP>
  • It requires the Ace DB database location, is none is supplied through the $ACEDB environment variable, it will search /usr for tace and use <it's grandparent directory>/wormbase
/usr/local/wormbase/wormmine/xmldumps/
[19:20|jdmswong@ip-10-35-66-254|acedb]$ ./imdump.sh /usr/local/wormbase/wormmine/xmldumps/
Did not specify AceDB dir in $ACEDB. Searching for tace...
ACEDB set to: /usr/local/wormbase/acedb/wormbase
Species
... done.
Gene
... done.
... and so on

If the directory passed into imdump.sh contains a trailing slash, script will not function correctly.

If XML files for each type in models has been created, then the script has executed successfully. The beginning and ends of each file may be checked manually to be sure.

If the error displayed in the output below is encountered, Ace is probably being disrupted and the command should be re-ran

jbaran@ip-10-35-66-254:~/src/website-intermine/acedb-dev/acedb$ ./imdump.sh /usr/local/wormbase/website/jbaran/acedbdump
Did not specify AceDB dir in $ACEDB. Searching for tace...
ACEDB set to: /usr/local/wormbase/acedb/wormbase
Species
... done.
Gene
... done.
Transcript
... done.
CDS
... done.
Variation
./imdump.sh: line 40: 19068 Killed                  $ACEDB_BIN/tace "$ACEDB"  > /dev/null <<EOF
wb

find ${model}
show -x -f "$dumpdir/$model.xml"
EOF

... done.
Manual queries

Not all records are desired for some types. In these cases specialized ace queries much be run, a step which is not yet automated.

Commands to generate WS239 build, in website-intermine/acedb-dev/acedb directory:

tace "/usr/local/wormbase/acedb/wormbase"

acedb> query find Gene Live
acedb> show -x -f <ACE XML DUMP>/Gene.xml

acedb> query find Protein Corresponding_CDS
acedb> show -x -f <ACE XML DUMP>/Protein.xml

acedb> query find CDS Method="curated"
acedb> show -x -f <ACE XML DUMP>/CDS.xml

acedb> query find Transcript (Gene)
acedb> show -x -f <ACE XML DUMP>/Transcript.xml

acedb> KeySet-Read species.ace
acedb> show -x -f <ACE XML DUMP>/Species.xml

The files generated will reflect the ace queries used to generate them. All species are loaded unless otherwise specified.

Each query in website-intermine/acedb-dev/acedb/manual_queries.txt must be run individually in tace, followed by show -x -f <TYPE>.xml where <TYPE> is the ace type being queried.

These queries represent desirable subsets of those represented types.

Acquire and pre-process data files

Copy the Ace XML over from the shared directory they were saved to from the machine which generated them or use:

rsync <HOST IP>:<HOST MACHINE FILE LOCATION> <TARGET MACHINE IP>:<TARGET MACHINE PATH>

Acquires data files from their appropriate data sources, and pre-processes each one accordingly.

On the InterMine machine, in the intermine directory:

  • Navigate to the redeployment folder.
intermine]$ cd redeployment/
  • The update.properties file should contain these two entries:
release = WS239
ace-xml-dir = /nfs/wormbase/wormmine/acedb_dumps/${release}

The release is used to generate strings, and must match the format of the FTP site. ace-xml-dir is where the build looks for the ace xml files generated above. This has to be set to <TARGET MACHINE PATH> from above.

The build downloads and processes fasta, gff3, go, gaf files. In addition to copying and processing the Ace XML files.

Build configuration
Property function
datadir The data directory for WormMine
release WormBase release version to use for paths and filenames
backup-dirname directory to backup old data directory too.
genomic-fasta-species-file species to download and/or process genomic fasta for
protein-fasta-species-file species to download and/or process protein fasta for
gff3-species-file same, for gff3s
ace-classes file ace classes to copy and/or process

ant -p will display all invokable tasks, available for individual execution.

  • Run the build

This will backup the old data directory into backup-dirname, delete it, then download and process all file types.

redeployment]$ ant

To only download and process:

redeployment]$ ant run-all

Size of datadir directories for WS239 build:

[12:24|jdmswong@wb-intermine-prod|datadir]$ du -sh *
2.7G    fasta
28M     go
28M     go-association
32G     wormbase-acedb
34G     wormbase-gff3

get-all-ace-xml:

BUILD FAILED
/home/jbaran/src/intermine/redeployment/build.xml:311: /home/jbaran/src/intermine/redeployment/ace_classes-test.txt doesn't exist

Total time: 85 minutes 15 seconds
Troubleshooting

Malformed JSON string: output produced by "ant run-all"

website-intermine/acedb-dev/intermine/redeployment$ ant run-all
Buildfile: /home/yourusername/src/intermine/redeployment/build.xml

get-assembly-ids:
     [exec] malformed JSON string, neither array, object, number, string or atom, at character offset 8424 (before "],\n      "full_name..."JSON retrieved from ftp://ftp.wormbase.org/pub/wormbase/releases/WS239/species/ASSEMBLIES.WS239.json
     [exec] ) at gen_assemblies.pl line 19.

BUILD FAILED
/home/yourusername/src/intermine/redeployment/build.xml:39: exec returned: 255

Total time: 2 seconds

Fix: Contact Kevin Howe so that the JSON file on the FTP server can be updated. JSON formatting errors occur sometimes when the file is manually edited after its generation.

Build the database

  • From the intermine directory, navigate to wormmine
cd wormmine
mkdir /YOURPATH/datadir/entrez-organism
  • Run build
wormmine]$ ../bio/scripts/project_build -b -v localhost wormmine_dump

This will run all sources configured in intermine/wormmine/project.xml file. To learn more about the project.xml file, refer to the official documentation

Any rare issues encountered can be addressed by the InterMine developer list at dev (AT) intermine.org

be sure to consult the log file

wormmine/pbuild.log

for details of any crashes which may occur.

Troubleshooting

If an exceptions occur, then "pbuild.log" should be checked for a more detailed explanation.

Build Steps (Preliminary Info)

Building the database refers to the process in which the project build script compiles each source together into a production database.

It is unclear which parts of the database are build using "build-db", which cannot recover/be resumed when it fails, and everything coming afterwards.

1. Build DB

Invoked by the -b switch to project_build.

Runs: cd dbmodel ; ant clean build-db

This generates the primary mine model file, and creates a fresh database with the desired schema. To generate the model it merged the model additions for all source types used in the main project.xml file. Ace XML, Gff3, and Fasta sources are the most essential.

Restarting Building from Milestones/Checkpoints

If a build error is encountered, it will appear in the output of the build command as "BUILD FAILED". To restart from the latest checkpoint, run the project build script with the -b (build) flag replaced with -l (recover). If the -b flag is not omitted, then "build-db" will be run again, which can take a long time.

   ../bio/scripts/project_build -v -l localhost wormmine_dump

This will, instead of rebuilding, attempt to restart by reading the last dump database. Dump databases are created by the build script from sources through specifications in the project.xml through the "dump" attribute. An example:

     <source name="wb-acedb-Variation" type="wormbase-acedb" dump="true">
          properties go here ...
     </source>

The builder will create the <DATABASE NAME>:wb-acedb-Variation database once this source is run. If restarting, the builder will find the most recent of these backup databases, clone it, and resume from there.

About the database

The database name is configured in the properties file as: db.production.datasource.databaseName. Each table represents a class in the model, with additional ones representing many-to-many collections, and various metadata. The InterMine development team does not currently advise for developers to modify the backend database due to many layers of inheritance, although questions may be directed to the InterMine developer list at dev (AT) intermine.org.

Instantiate database dump

Backup archive can be found at

/nfs/wormbase/wormmine/database_dumps

To instantiate a previously built WormMine production database.

  • Find your favorite release from WORMMINE DB FTP URL (placeholder, no URL exists)
  • Create empty DB
> createdb -U intermine -E SQL_ASCII wormmine
    • -U: user set to intermine
    • -E: character set used
  • Unpack and restore DB
> psql -U intermine -d wormmine -f <WORMMINE RELEASE SQL>
    • -U: execute as user
    • -d: destination DB
    • -f: SQL input file

Copy an existing database

Creates a new database with the contents of an already existing database:

createdb -U intermine -W -T EXISTINGDBNAME NEWDBNAME

Note: the existing database needs to be owned by the "intermine" PostgreSQL user. If that is not the case, then it can be set using the following SQL command:

ALTER DATABASE EXISTINGDBNAME OWNER TO intermine

Migrating from the unmerged to staging branch

git merge is used to move the files and changes on the unmerged branch into staging.

[15:32|jdmswong@wb-intermine|webapp]$ git checkout staging
Switched to branch 'staging'
[15:34|jdmswong@wb-intermine|webapp]$ git merge unmerged --squash
.......
[15:36|jdmswong@wb-intermine|webapp]$ git commit

Migrating to production machine

The production database must be moved to the production server to be served by the webapp.

Any paths used in the following commands are arbitrary as long as they are consistent.

Some databases are named along the lines of wormmine-ws239-2. Database names are arbitrary as long as they are referred to correctly in the mine properties file.

Dumping database

On the development machine:

pg_dump -U intermine <DATABASE NAME> -f <PATH TO DUMPFILE>

This created the database dump file.

Transfer this dumpfile to the production machine. Scp is one of the many options: On development machine:

scp <PATH TO DUMPFILE> <REMOTE MACHINE IP>:<REMOTE DUMPFILE PATH>

Restoring database

On production machine:

createdb -U intermine -E SQL_ASCII <DESTINATION DATABASE NAME>
psql -U intermine -d <DESTINATION DATABASE NAME> -f <REMOTE DUMPFILE PATH>

The database is now instantiated on the production machine

If the following error is encountered, your postgres user must be granted super user privilege.

jbaran@wb-intermine:/nfs/wormbase/archive/jbaran/WS239-2$ psql -U intermine -d joachim-ws239-2 -f wormmine-ws239-2.sql
SET
SET
SET
SET
SET
SET
SET
psql:wormmine-ws239-2.sql:18: ERROR:  must be superuser to create a base type
psql:wormmine-ws239-2.sql:27: ERROR:  permission denied for language c
ALTER FUNCTION
psql:wormmine-ws239-2.sql:38: ERROR:  permission denied for language c
ALTER FUNCTION
psql:wormmine-ws239-2.sql:53: ERROR:  must be superuser to create a base type
ALTER TYPE
psql:wormmine-ws239-2.sql:62: ERROR:  must be owner of type bioseg
psql:wormmine-ws239-2.sql:71: ERROR:  permission denied for language c
ALTER FUNCTION
psql:wormmine-ws239-2.sql:80: ERROR:  must be owner of function bioseg_cmp
psql:wormmine-ws239-2.sql:89: ERROR:  permission denied for language c
ALTER FUNCTION
psql:wormmine-ws239-2.sql:98: ERROR:  must be owner of function bioseg_contained
psql:wormmine-ws239-2.sql:107: ERROR:  permission denied for language c
ALTER FUNCTION
psql:wormmine-ws239-2.sql:116: ERROR:  must be owner of function bioseg_contains
psql:wormmine-ws239-2.sql:125: ERROR:  permission denied for language c

Configuring mine to database

On production machine, in

~/.intermine/wormmine.properties

, set

db.production.datasource.databaseName=<DESTINATION DATABASE NAME> 

Create userprofile database

InterMine needs a separate database to track users and their information.

This can be skipped if an existing userprofile database is present.

  • Create empty DB
> createdb -U intermine -E SQL_ASCII userprofile-wormmine
  • Build the userprofile DB
> cd wormmine/webapp
> ant build-db-userprofile

This formats the empty userprofile database for mine use.

About the userprofile database

The database name is set in the properties file as db.userprofile-production.datasource.databaseName. User information is stored in the userprofile table. Tables that begin with "saved" map users to any data they have saved; such as lists, queries, templates, and so on. List data mapping is stores in bagvalues

Launch webapp

  • Note: Catalina must be running before you deploy the webapp (about catalina)
  • Navigate to intermine/wormmine/webapp
  • Launch webapp:
 > ./xx

This script contains:

ant clean
ant -v default remove-webapp release-webapp

Which may be run in sequence instead. These commands clear previous webapp files, remove any existing webapps which may be launched, and compile and release a new webapp.

Test Webapp

You should be able to reach your new instance through <baseurl>/wormmine Webapp is standalone.

About Catalina

Catalina is Tomcat's servlet container. Catalina implements specs for servlet and JSP.

$CATALINA_HOME = /home/jdmswong/website-intermine/software/tomcat/apache-tomcat-6.0.36


Catalina must be running before you deploy the webapp

$CATALINA_HOME/bin/startup.sh


In case of problems with deploying the webapp, try restarting Catalina

$CATALINA_HOME/bin/shutdown.sh
$CATALINA_HOME/bin/startup.sh

Logging

Logs can be found:

intermine/wormmine/webapp/intermine.log
<$CATALINA_HOME>/logs

Note:

$CATALINA_HOME = /home/jdmswong/website-intermine/software/tomcat/apache-tomcat-6.0.36

TODO: document which logs record what

Attach to WormBase instance

If you want to enable integration with WormBase, follow these steps:

Checkout merged branch

> git checkout remotes/origin/staging
Note: checking out 'remotes/origin/staging'.

You are in 'detached HEAD' state. You can look around, make experimental
changes and commit them, and you can discard any commits you make in this
state without impacting any branches by performing another checkout.

If you want to create a new branch to retain commits you create, you may
do so (now or later) by using -b with the checkout command again. Example:

  git checkout -b new_branch_name

HEAD is now at 611b791... new tests, shell script to run tests

> git checkout -b staging
Switched to a new branch 'staging'

Reconfigure properties file

Properties file located ~/.intermine/wormmine.properties (currently /home/jdmswong/.intermine/wormmine.properties/)

  • Needs to deploy at tools/wormmine
webapp.path=tools/wormmine
  • Update url as appropriate:
  • For deployment on staging.wormbase.org:
webapp.baseurl=http://staging.wormbase.org
webapp.returnurl=http://staging.wormbase.org/auth/openid?openid_identifier=https://www.google.com/accounts/o8/id&redirect=http://dev.wormbase.org/tools/wormmine/mymine.do#

project.sitePrefix=http://staging.wormbase.org/tools/wormmine

  • For deployment at www.wormbase.org
webapp.baseurl=http://www.wormbase.org
webapp.returnurl=http://www.wormbase.org/auth/openid?openid_identifier=https://www.google.com/accounts/o8/id&redirect=http://www.wormbase.org/tools/wormmine/mymine.do#

project.sitePrefix=http://www.wormbase.org/tools/wormmine

Modify wormbase.conf

To enable login system, make sure config flag: wormmine_path = 'tools/wormmine' is uncommented.

Upgrading between releases

This process has not been fully automated yet, and thus requires some manual work.

Upgrade release database

Instructions above

Upgrade wormmine.properties file

The current release uses /home/jdmswong/.intermine/wormmine.properties.

Relevant properties:

# this is the production database the mine will use
db.production.datasource.databaseName=wormmine-ws238-3

# This appears at the top next to Version WS
project.releaseVersion= 238 IM v1.2.1

Update genomic_model.xml

This is the central model file, used by the webapp process. It must be imported from the machine which produced the production database.

Note: it can be added to the repository, but updates after each test build. This led to extraneous commits and merging conflicts.

In /home/jdmswong/intermine/wormmine/dbmodel/build/model

[17:44|jdmswong@wb-intermine|model]$ rsync 206.108.125.174:/home/jdmswong/idev/wormmine/dbmodel/build/model/genomic_model.xml .

Development in a Nutshell

Please provide a walkthrough on how new AceDB class could be added to WormMine. List which files need to be added (incl. why), which configuration files need to be updated, and how https://github.com/WormBase/intermine/blob/unmerged/build_config/wormbase-acedb/Gene_mapping.properties mappings work.

Property files are projecting columns (?) to Ace values via XPath expressions. It is unclear where the data ends up though.