Skip to content

Commit

Permalink
srm: Drop Terracotta support
Browse files Browse the repository at this point in the history 
Target: trunk
Require-notes: yes
Require-book: no
Acked-by: Paul Millar <paul.millar@desy.de>
Acked-by: Dmitry Litvintsev <litvinse@fnal.gov>
Patch: https://rb.dcache.org/r/8207/
  • Loading branch information
@gbehrmann
gbehrmann committed May 13, 2015
1 parent f7033ea commit b97a41f
Show file tree
Hide file tree
Showing 4 changed files with 15 additions and 371 deletions.
91 changes: 15 additions & 76 deletions modules/srm-server/src/main/java/org/dcache/srm/request/Job.java
View file
Expand Up @@ -365,44 +365,19 @@ private boolean isValidTransition(State currentState, State newState)
}

/**
* Try to change the state of the job to READY. the request into the ready state
* Try to change the state of the job into the READY state.
*/
public void tryToReady() {
rlock();
wlock();
try {
if(state != State.RQUEUED) {
return;
if (state == State.RQUEUED && schedulerId != null) {
Scheduler scheduler = Scheduler.getScheduler(schedulerId);
if (scheduler != null) {
scheduler.tryToReadyJob(this);
}
}
} finally {
runlock();
}

/*
* The job should be readied, only if the job's scheduler's
* count of the "ready" jobs is bellow the maximum allowed number of the
* Ready jobs. If the job is replicated in multiple jvm's,
* the job might not have a scheduler associated with it in this jvm,
* then in this jvm nothing needs to happen. Whatever clastering
* mechanizm used needs to makes sure that if this method is called in
* an instance that is different from the one where the job was
* originally scheduled leads to the call of the invocation of the same
* method in the instance of jvm where the job was orignally
* scheduled, and where the scheduler is non-null.
* In case of terracotta it is achived by including this method in the
* "distributed-method" section of the configuration file :
* <distributed-methods>
* <method-expression>
* void Job.tryToReady()
* </method-expression>
* <method-expression>
*/
if(schedulerId != null) {
Scheduler scheduler = Scheduler.getScheduler(schedulerId);
// we excpect that the scheduler in not null only in the jvm
// that created and scheduled the job in the first place
if(scheduler != null) {
scheduler.tryToReadyJob(this);
}
wunlock();
}
}

Expand Down Expand Up @@ -961,51 +936,15 @@ public void scheduleWith(Scheduler scheduler) throws InterruptedException,
* @param newState
*/
private void notifySchedulerOfStateChange(State oldState, State newState) {
// we need to lock again, even if the call is from the setState method
// where weh have locked already. The method is configured to run in
// multiple jvms though "dsitributed method" invocation, which means
// the method can be called on its own by terracotta framework
wlock();
try {
/*
* The state change needs to be correctly accounted by the scheduler that
* executes this job. This is done by call to scheduler's stateChanged
* method. If the job is replicated in multiple jvm's, the job might have
* a scheduler associated with it in different jvm.
* Then in this jvm call to stateChanged should not take place.
* Insted whatever clastering mechanizm is used needs to makes sure that
* if this method is also called in an instance where the job was
* originally scheduled and where the Scheduler.getScheduler(schedulerId)
* will return non null result.
* In case of terracotta it is achived by including this method in the
* "distributed-method" section of the configuration file :
* <distributed-methods>
* <method-expression>
* void Job.notifySchedulerOfStateChange(..)
* </method-expression>
* <method-expression>
*
*/

if (schedulerId != null) {
Scheduler scheduler = Scheduler.getScheduler(schedulerId);
// we excpect that the scheduler in not null only in the jvm
// that created and scheduled the job in the first place
if (scheduler != null) {
// code bellow is executed only in one SRM in the cluster
// the SRM that created this job
logger.debug("notifySchedulerOfStateChange calls " +
"scheduler.stateChanged()");
scheduler.stateChanged(this, oldState, newState);
// set schedulerId to null only in jvm
// that has scheduler that "owns" this job
if(state.isFinal()) {
schedulerId = null;
}
if (schedulerId != null) {
Scheduler scheduler = Scheduler.getScheduler(schedulerId);
if (scheduler != null) {
logger.debug("notifySchedulerOfStateChange calls scheduler.stateChanged()");
scheduler.stateChanged(this, oldState, newState);
if (state.isFinal()) {
schedulerId = null;
}
}
} finally {
wunlock();
}
}

Expand Down
253 changes: 0 additions & 253 deletions skel/etc/tc-config.xml
View file

This file was deleted.

23 changes: 0 additions & 23 deletions skel/share/defaults/dcache.properties
View file
Expand Up @@ -223,29 +223,6 @@ dcache.log.access.max-history=30
# if any of the input files have changed.
(not-for-services)dcache.config.cache=true

# The following property describes whether dCache should run under Terracotta.
# It is only supported by the srm service at this time,so do not enable it in
# dcache.conf; instead, enable it, in the layout file, for the domain the srm
# service runs within.
#
# For example:
#
# [srmDomain]
# dcache.terracotta.enabled=true
# dcache.terracotta.install.dir=/opt/terracotta
#
# [srmDomain/srm]
# [srmDomain/spacemanager]
#
(not-for-services,one-of?true|false)dcache.terracotta.enabled=false

# The following parameter specifies the location of Terracotta
# If dcache.terracotta.enabled is true then this must be specified as well
(not-for-services)dcache.terracotta.install.dir=

# Location of the Terracotta configuration file
(not-for-services)dcache.terracotta.config.path=${dcache.paths.etc}/tc-config.xml

# -----------------------------------------------------------------------
# Parameters related to what runs inside a domain
# -----------------------------------------------------------------------
Expand Down

0 comments on commit b97a41f

Please sign in to comment.