$Id: RELEASE_NOTES 1795 2009-04-22 21:06:21Z vahi $
===============================
Release Notes for PEGASUS 2.3.0
===============================
NEW FEATURES
--------------
1) Regex Based Replica Selection
Pegasus now allows users to use regular expression based replica
selection. To use this replica selector, users need to set the
following property
pegasus.selector.replica Regex
The Regex replica selector allows the user allows the user to
specifiy the regex expressions to use for ranking various PFNs
returned from the Replica Catalog for a particular LFN. This
replica selector selects the highest ranked PFN i.e the replica
with the lowest rank value.
The regular expressions are assigned different rank, that determine
the order in which the expressions are employed. The rank values
for the regex can expressed in user properties using the property.
pegasus.selector.replica.regex.rank.[value]
The value is an integer value that denotes the rank of an
expression with a rank value of 1 being the highest rank.
For example, a user can specify the following regex expressions
that will ask Pegasus to prefer file URL's over gsiftp url's from
example.isi.edu
pegasus.selector.replica.regex.rank.1 file://.*
pegasus.selector.replica.regex.rank.2 gsiftp://example\.isi\.edu.*
User can specify as many regex expressions as they want.
Since Pegasus is in Java , the regex expression support is what
Java supports. It is pretty close to what is supported by
Perl. More details can be found at
http://java.sun.com/j2se/1.5.0/docs/api/java/util/regex/Pattern.html
There is documentation about the new replica selector in the
properties document . It can also be found at
$PEGASUS_HOME/etc/sample.properties
To use this set pegasus.selector.replica Regex
2) Automatic Determination of pool attributes in RLS Replica Catalog
Pegasus can now associate a pool attribute with the replica catalog
entries returned from querying a LRC if the pool attribute is not
already specified.
This is achieved by associating the site handles with corresponding
LRC url's in the properties file. This mapping tells us what
default pool attribute should be assigned while querying a
particular LRC. For example
pegasus.catalog.replica.lrc.site.llo rls://ldas.ligo-la.caltech.edu:39281
pegasus.catalog.replica.lrc.site.lho rls://ldas.ligo-wa.caltech.edu:39281
tells Pegasus that all results from LRC
rls://ldas.ligo-la.caltech.edu:39281 are associated with site llo
Using this feature only makes sense, when a LRC *ONLY* contains
mapping for data on one site, as in case of LIGO LDR deployment.
3) Pegasus auxillary jobs on submit host now execute in local universe
All the scheduler universe jobs are now executed in local
universe. Also any job planned for site local will by default run
in local universe instead of scheduler universe.
Additionally, extra checks were put in to handle the Condor File
Transfer Mechansim issues in case local/scheduler universe. This
was tracked in bugzilla at
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=40
A user can override the local universe generation by specifying the
condor profile key universe and setting it to the value desired.
4) Python API for generating DAX and PDAX
Pegasus now includes a Python API for generating DAXes and PDAXes.
An example can be found online at
http://vtcpc.isi.edu/pegasus/index.php/ChangeLog#Added_Python_API_for_DAX_and_PDAX
For more information on the DAX API type: pydoc Pegasus.DAX2
For more information on the PDAX API type: pydoc Pegasus.PDAX2
5) Interface to Engage VO for OSG
There is a new Site Catalog Implementation called Engage that
interfaces with the Engage VO to discover resource information
about OSG from the information published in RENCI glue classads.
To use it set
pegasus.catalog.site Engage
To generate a site catalog using pegasus-get-sites set the source
option to Engage
pegasus-get-sites --source Engage --sc engage.sc.xml
6) Gensim now reports Seqexec Times and Seqexec Delays
Gensim script ($PEGASUS_HOME/contrib/showlog/gensim) now reports
the seqexec time and the seqexec delay for the clustered jobs.
There are two new columns in the jobs file created by seqexec
- seqexec
- seqexec delay.
The seqexec time is determined from the last line of the .out file
of the clustered jobs. E.g format [struct stat="OK", lines=4,
count=4, failed=0,
duration=21.836,start="2009-02-20T16:14:56-08:00"]
The seqexec delay is the seqexec time - kickstart time.
This useful for analyzing large scale workflow runs.
7) Properties to turn on or off the seqexec progress logging
The property pegasus.clusterer.job.aggregator.seqexec.hasgloballog
is now deprecated.
It has been replaced by two boolean properties
- pegasus.clusterer.job.aggregator.seqexec.log whether to log
progress or not
- pegasus.clusterer.job.aggregator.seqexec.log.global whether to
log progress to global file or not.
The pegasus.clusterer.job.aggregator.seqexec.log.global only
comes into effect when
pegasus.clusterer.job.aggregator.seqexec.log is set to true
8) Passing of the DAX label to kickstart invocation
Now, the kickstart invocation for the jobs is always passed the dax
label using the -L option. To disable the passing of the DAX label,
user needs to set pegasus.gridstart.label to false
Additionally, the basename option to pegasus-plan overrides the
label value retrieved from the DAX.
9) show-job works on MAC OSX platform
$PEGASUS_HOME/contrib/showlog/show-job now does not fail on
unavailability of convert program. It only logs a warning and
creates the EPS File , but not the png files. This allows us to
run show-job on MAC OSX systems.
10) Enabling InPlace cleanup in deferred planning
By default in case of deferred planning cleanup is turned off as
the cleanup algorithm does not work across partitions.
However, in scenarios where the partitions themseleves are
independant ( i.e. dont share files ), user can safely turn on
cleanup.
This can now be done by setting
pegasus.file.cleanup.scope deferred
If the property is set to deferred, and the users wants to disable
cleanup , they can still specify --nocleanup option on command
line and that is honored.
However in case of scope fullahead for deferred planning, the
command line options are ignored and always nocleanup is set.
11) New Pegasus Job Classad
Pegasus now publishes a job runtime classad with the jobs. The
class ad key name is pegasus_job_runtime. The value passed to it
is picked up from the Pegasus Profile runtime. If the Pegaus
Profile is not associated, then the globus maxwalltime profile key
is used. If both are not set, then a value of zero is published.
This job classad can be used for users in case of glidein, to
ensure that the jobs complete before the nodes expire.
For the coral glidein service the sub expression to job
requirement swould look something like this
(CorralTimeLeft > MY.pegasus_job_runtime)
12) [workflow].job.map file
Pegasus now creates a [workflow].job.map file that links jobs in
the DAG with the jobs in the DAX. The contents of the file are in
netlogger format.
The [workflow] is replaced by the name of the workflow i.e. same
prefix as the .dag file
In the file there are two types of events.
a) pegasus.job
b) pegasus.job.map
pegasus.job - This event is for all the jobs in the DAG. The
following information is associated with this event.
- job.id the id of the job in the DAG
- job.class an integer designating the type of the job
- job.xform the logical transformation which the job refers to.
- task.count the number of tasks associated with the job. This is
equal to the number of pegasus.job.task events created for that
job.
pegasus.job.map - This event allows us to associate a job in the
DAG with the jobs in the DAX. The following information is
associated with this event.
-task.id the id of the job in the DAG
-task.class an integer designating the type of the job
-task.xform the logical transformation which the job refers to.
13) Source Directory for Worker Package Staging
Users now can specify the property
pegasus.transfer.setup.source.base.url to specify the URL to the
source directory containing the pegasus worker packages. If it is
not specified, then the worker packages are pulled from the http
server at pegasus.isi.edu during staging of executables.
BUGS FIXED
----------
1) Critical Bug Fix to rc-client
SCEC reported a bug with the rc-client while doing bulk inserts
into RLS. The bug was related to how logging is initialized
internally in the client.
Details of the bug fix can be found at
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=38
2) Bug Fix to tailstatd for parsing jobnames with . in them
There was a bug where tailstatd incorrectly generated events in
the jobstate.log while parsing condor logs. This was due to an
errorneous regex expression for determining the event
POST|PRE SCRIPT STARTED.
The earlier expression did not allow for . in jobnames. This is
especially prevalent in LIGO workflows where the DAX labels have
. in them.
An example of the problem line in DAGMan log
1/24 10:11:21 Running POST script of Node
inspiral_hipe_eobinj_cat2_veto.EOBINJ_CAT_2_VETO.daxlalapps_sire_ID000731...
Earlier the job id was parsed as inspiral_hipe_eobinj_cat2_veto
instead of
inspiral_hipe_eobinj_cat2_veto.EOBINJ_CAT_2_VETO.daxlalapps_sire_ID000731
3) Pegasus Builds on FC10
Earlier the Pegasus builds were failed on FC10 as the invoke c tool
did not build correctly. This is now fixed.
Details at
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=41
4) tailstatd killing jobs by detecting starvation
tailstatd removes a job after four hours when the job has been
waiting in the queue WITHOUT being marked as EXECUTE in the condor
log. To override tailstatd has an option of setting starvation time
to 0 via command line or via pegasus.max.idletime property. The if
condition in the perl script was not accepting 0 as a value when
trying to override the default 4 hour starvation time. This fix
allows the value to be set to 0 (turn of starvation checks) or any
other value via the property pegasus.max.idletime.
This was tracked in pegasus jira as bug 40
http://pegasus.isi.edu/jira/browse/PM-40
Documentation
--------------
1) User Guides
The release has new user guides about the following
- Pegasus Job Clustering
- Pegasus Profiles
- Pegasus Replica Selection
The guides are checked in $PEGASUS_HOME/doc/guides
They can be found online at
http://pegasus.isi.edu/mapper/doc.php
2) Property Document was updated with the new properties introduced.
===============================
Release Notes for PEGASUS 2.2.0
===============================
NEW FEATURES
--------------
1) Naming scheme changed for auxillary jobs
Pegasus during the refinement of the abstract workflow to the
executable workflows adds auxillary jobs to do data stagein/stageout,
create work directories for workflow etc. The prefixes/suffixes added
for these jobs has been changed.
Type of Job | Old Prefix | New Prefix
-------------------------------------------------------------------
Data Stage In Job | rc_tx_ | stage_in_
Data Stage Out Job | new_rc_tx_ | stage_out_
Data Stage In Job between sites | inter_tx_ | stage_inter_
Data Registration Job | new_rc_register_ | register_
Cleanup Job | cln_ | clean_up_
Transfer job to transfer the | setup_tx_ | stage_worker_
worker package | |
Additionally, the suffixes for the create dir jobs are now replaced
by prefixes
Type of Job | Old Suffix | New Prefix
-------------------------------------------------------------------
Directory creation job | _cdir | create_dir_
Synch Job in HourGlass mode | pegasus_concat | pegasus_concat_
2) Staging of worker package to remote sites
Pegasus now supports staging of worker package as part of the workflow.
This feature is tracked through pegasus bugzilla .
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=35
The worker package is staged automatically to the remote site, by
adding a setup transfer job to the workflow.
The setup transfer job by default uses GUC to stage the data.
However, this can be configured by setting the property
pegasus.transfer.setup.impl property. If you also, have
pegasus.transfer.*.impl set in your properties file, then you need
explicilty set pegasus.transfer.setup.impl to GUC
The code discovers the worker package by looking up pegasus::worker
in the transformation catalog.
Note: that the basename of the url's should not be changed. Pegasus
parses the basename to determine the version of the worker package.
Pegasus automatically determines the location of the worker package
to deploy on the remote site. Currently default mappings are as
follows
INTEL32 => x86
AMD64 => x86_64 or x86 if not available
INTEL64 =>x86
OS LINUX = rhel3
There is an untar job added to the workflow after the setup job that
un tars the worker package on the remote site. It defaults to /bin/tar .
However can be overriden by specifying the entry tar in the
transformation catalog for a particular site.
3) New Site Catalog Schema
This release of Pegasus has support for site catalog schema version 3.
HTML visualization of schema:
http://pegasus.isi.edu/mapper/docs/schemas/sc-3.0/sc-3.0.html
Schema itself:
http://pegasus.isi.edu/schema/sc-3.0.xsd
A sample xml file :
http://pegasus.isi.edu/schema/sc-3.0-sample.xml
To use a site catalog in the new format set
pegasus.catalog.site XML3
Changes to sc-client
sc-client command line tool was updated to convert an existing
site catalog from old format to the new format.
Sample usage
sc-client -i ldg-old-sites.xml -I XML -o ldg-new-sites.xml -O XML3
sc-client --help gives detailed help
4) pegasus-get-sites
pegasus-get-sites was recoded in JAVA and now generates the site
catalog confromant to schema sc-3.0.
Sample Usage to query VORS to generate a site catalog for OSG.
pegasus-get-sites --source VORS --grid osg -s ./sites-new.xml
The value passed to the source option is case sensitive.
Additionally, the VORS module of pegasus-get-sites determines the
value of GLOBUS_LOCATION variable dependant on whether the
auxillary jobmanager is of type fork or not.
If it is of type fork then picks up the value of GLOBUS_LOCATION
variable published in VORS for that site. else it picks up the
value from OSG_GRID variable published in VORS for that
site. i.e. GLOBUS_LOCATION is set to $OSG_GRID/globus
5) Overhaul of logging
The Pegasus logging interfaces have been reworked.
Now users can specify the logger they want to use, by specifying the
property pegasus.log.manager .
Currently, two logging implementations are supported.
Default - Pegasus homegrown logger that logs to stdout and stderr
directly.
Log4j - Uses log4j to log the messages.
The Log4j properties can be specified at runtime by specifying the
property pegasus.log.manager.log4j.conf
The format of the log message themselves can be specified at runtime
by specifying the property pegasus.log.manager.formatter
Right now two formatting modes are supported
a) Simple - This formats the messages in a simple format. The
messages are logged as is with minimal formatting. Below are
sample log messages in this format while ranking a dax according
to performance.
event.pegasus.ranking dax.id se18-gda.dax - STARTED
event.pegasus.parsing.dax dax.id se18-gda-nested.dax - STARTED
event.pegasus.parsing.dax dax.id se18-gda-nested.dax - FINISHED
job.id jobGDA
job.id jobGDA query.name getpredicted performace time 10.00
event.pegasus.ranking dax.id se18-gda.dax - FINISHED
b) Netlogger - This formats the messages in the Netlogger format ,
that is based on key value pairs. The netlogger format is useful
for loading the logs into a database to do some meaningful analysis.
Below are sample log messages in this format while ranking a dax
according to performance.
ts=2008-09-06T12:26:20.100502Z event=event.pegasus.ranking.start \
msgid=6bc49c1f-112e-4cdb-af54-3e0afb5d593c \
eventId=event.pegasus.ranking_8d7c0a3c-9271-4c9c-a0f2-1fb57c6394d5 \
dax.id=se18-gda.dax prog=Pegasus
ts=2008-09-06T12:26:20.100750Z event=event.pegasus.parsing.dax.start \
msgid=fed3ebdf-68e6-4711-8224-a16bb1ad2969 \
eventId=event.pegasus.parsing.dax_887134a8-39cb-40f1-b11c-b49def0c5232\
dax.id=se18-gda-nested.dax prog=Pegasus
ts=2008-09-06T12:26:20.100894Z event=event.pegasus.parsing.dax.end \
msgid=a81e92ba-27df-451f-bb2b-b60d232ed1ad \
eventId=event.pegasus.parsing.dax_887134a8-39cb-40f1-b11c-b49def0c5232
ts=2008-09-06T12:26:20.100395Z event=event.pegasus.ranking \
msgid=4dcecb68-74fe-4fd5-aa9e-ea1cee88727d \
eventId=event.pegasus.ranking_8d7c0a3c-9271-4c9c-a0f2-1fb57c6394d5 \
job.id="jobGDA"
ts=2008-09-06T12:26:20.100395Z event=event.pegasus.ranking \
msgid=4dcecb68-74fe-4fd5-aa9e-ea1cee88727d \
eventId=event.pegasus.ranking_8d7c0a3c-9271-4c9c-a0f2-1fb57c6394d5 \
job.id="jobGDA" query.name="getpredicted performace" time="10.00"
ts=2008-09-06T12:26:20.102003Z event=event.pegasus.ranking.end \
msgid=31f50f39-efe2-47fc-9f4c-07121280cd64 \
eventId=event.pegasus.ranking_8d7c0a3c-9271-4c9c-a0f2-1fb57c6394d5
6) New Transfer Refiner
Pegasus has a new transfer refiner named Cluster.
In this refinement strategy, clusters of stage-in and stageout jobs are
created per level of the workflow. It builds upon the Bundle refiner.
The differences between the Bundle and Cluster refiner are as follows.
- stagein is also clustered/bundled per level. In Bundle it was
for the whole workflow.
- keys that control the clustering ( old name bundling are )
cluster.stagein and cluster.stageout instead of bundle.stagein and
bundle.stageout
This refinement strategy also adds dependencies between the stagein
transfer jobs on different levels of the workflow to ensure that stagein
for the top level happens first and so on.
An image of the workflow with this refinement strategy can be found
at
http://vtcpc.isi.edu/pegasus/index.php/ChangeLog#Added_a_Cluster_Transfer_Refiner
7) New Transfer Implementation for GUC from globus 4.x
Pegasus has a new transfer implementation that allows it to use GUC
from globus 4.x series to transfer multiple files in one job.
In order to use this transfer implementation
- the property pegasus.transfer.*.impl must be set to value GUC.
There should be an entry in the transformation catalog with the
fully qualified name as globus::guc for all the sites where
workflow is run, or on the local site in case of third party
transfers.
Pegasus can automatically construct the path to the globus-url-copy
client, if the environment variable GLOBUS_LOCATION is specified in
the site catalog for the site.
The arguments with which the client is invoked can be specified
- by specifying the property pegasus.transfer.arguments
- associating the Pegasus profile key transfer.arguments
8) Recursive DAX'es
There is prototypical support for recursive dax'es. Recursive
DAX'es give you the ability to specify a job in the DAX that points
to another DAX that has to be executed.
There is a sample recursive dax at
$PEGASUS_HOME/examples/recursive.dax
The dax refers to pegasus jobs in turn plan and execute a dax
To get this dax planned by pegasus you will need to have additional
entries for dagman and pegasus in your transformation catalog.
For e.g.
local condor::dagman /opt/condor/7.1.0/bin/condor_dagman INSTALLED INTEL32::LINUX NULL
local pegasus::pegasus-plan:2.0 /lfs1/software/install/pegasus/default INSTALLED INTEL32::LINUX NULL
The recursive dax needs to be planned for site local, since the
pegasus itself runs on local site. The jobs in the dax specify -s
option where you want each of your workflows to run.
Recursive DAX do not need to contain only pegasus jobs. They can
contain application/normal jobs that one usually specifies in a
DAX. Pegasus determines that a particular job is planning and
execute job by looking for a pegasus profile key named type with
value recursive e.g.
recursive-Dpegasus.user.properties=/lfs1/work/conf/properties
--dax /lfs1/work/dax3 -s tacc -o local --nocleanup --force
--rescue 1 --cluster horizontal -vvvvv --dir ./dag_3
09) Rescue option to pegasus-plan for deferred planning
A rescue option to pegasus-plan has been added. The rescue option
takes in an integer value, that determines the number of times
rescue dags are submitted before re-planning is triggered in case
of failures in deferred planning. For this to work, Condor 7.1.0
or higher is required as it relies on the recently implemented
auto rescue feature in Condor DAGMan.
Even though re-planning is triggered, Condor DAGMan still ends up
submitting the rescue dag as it auto detects. The fix to it is to
remove the rescue dag files in case of re-planning. This is still
to be implemented
10) -j|--job-prefix option to pegasus-plan
pegasus-plan can now be passed the -j|--job-prefix option to
designate the prefix that needs to be used for constructing the
job submit file.
11) Executing workflows on Amazon EC2
Pegasus now has support of running workflows on EC2 with the
storage of files on S3. This feature is still in testing phase and
has not been tested fully.
To execute workflows on EC2/S3, Pegasus needs to be configured to
use S3 specific implementations of it's internal API's
a) First level Staging API - The S3 implementation stages in from the
local site ( submit node ) to a bucket on S3. Similarly the data is
staged back from the bucket to the local site ( submit node )
. All the first level transfers happen between the submit node
and the cloud. This means that input data can *only* be present
on the submit node when running on the cloud, and the output
data can be staged back only to the submit node.
b) Second Level Staging API - The S3 implementation retrieves input
data from the bucket to the worker node tmp directory and puts
created data back in the bucket.
c) Directory creation API - The S3 implementation creates a bucket
on S3 for the workflow instead of a directory.
d) Cleanup API - To cleanup files from the workflow specific bucket
on S3 during workflow execution.
The above implementations rely on s3cmd command line client to
interface with S3 filesystem. There should be an entry in the
transformation catalog with the fully qualified name as
amazon::s3cmd for the site corresponding to the cloud and the local
site.
To configure Pegasus to use these implementations set the following
properties
pegasus.transfer.*.impl S3
pegasus.transfer.sls.*.impl S3
pegasus.dir.create.impl S3
pegasus.file.cleanup.impl S3
pegasus.execute.*.filesystem.local true
12) Support for OSU Datacutter jobs
Pegasus has new gridstart mode called DCLauncher. This allows us to
launch the Data Cutter jobs using the wrapper that OSU group wrote.
Pegasus now supports the condor parallel universe.
To launch a job using DCLauncher, the following pegasus profile
keys need to be associated with the job
gridstart to DCLauncher
gridstart.path the path to the DCLauncher script
13) New Pegasus Profiles Keys
a) create.dir - this profile key triggers kicstart to create and
change directories before launching a job.
b) gridstart.path - this profile key specifies the path to the
gridstart used to launch a particular job
c) runtime - this profile key is useful when using Heft based site
selection. It allows users to associate expected runtimes of
jobs with the job description in DAX.
14) Kickstart captures machine information
Kickstart now logs machine information in the invocation record
that it creates for each job invocation. The Kickstart JAVA parser
can parse both records in old and new format.
A snippet of machine information captured is show below
2008-09-23T13:58:05.211-07:00#2 SMP Thu Apr 28 18:41:14 PDT 20052008-09-12T12:03:49.772-07:00Intel(R)
Xeon(TM) CPU 2.40GHz
15) Kickstart works in cygwin environment
Kickstart now compiles on cygwin. Kickstart could not find
SYS_NMLN variable in Cygwin to determine the uname datastructure's
size. Added a fix in the Makefile to add CFLAGS -DSYS_NMLN=20 when
the OS is Cygwin/Windows
The kickstart records generated on cygwin are slightly different
from the ones generated unix platforms. The kickstart parser was
modified to handle that.
The differences are as follows -
a) On cygwin inode value is double. The inode value is parsed as
double , but cast to long to prevent errors.
b) On cygwin the uid and gid values are long. They are passed as
long, but cast to int to prevent errors.
16) Changes to dirmanager
The dirmanager executable can now remove and create multiple
directories. This is achieved by specifying a whitespace separated
list of directories to the --dir option.
17) Added color-file option to showjob
There is now a --color-file option to show-job in
$PEGASUS_HOME/contrib/showlog to pass a file that has the mappings
from transformation name to colors.
The format of each line is as follows
transformation-name color
This can be used to assign different colors to compute jobs in a
workflow. The default color assigned is gray if none is
specified.
18) jobstate-summary tool
There is a new tool at $PEGASUS_HOME/bin/jobstate-summary.
It attempts to give a summary for the workflow. Should help in
jobstate-summ debugging failed job information. It will shows all
the information associated with a failed job. It gets the list of
failed job from the jobstate.log file. After that it parses latest
kickstart file for each failed job and show the exit code and all
the other information.
Usage: jobstate-summary --i [--v(erbose)]
[--V(ersion)] [--h(elp)]
Input directory is the place where all the log files including jobstate.log file reside.
v option is for verbose debugging.
V option gives the pegasus version.
h option prints the help message.
A sample run is like jobstate-summary -i /dags/pegasus/diamond/run0013 -v
19) Support for DAGMan node categories
Pegasus now supports DAGMan node categories. DAGMan now allows to
specify CATEGORIES for jobs, and then specify tuning parameters (
like maxjobs ) per category. This functionality is exposed in
Pegasus as follows
The user can associate a dagman profile key category with the
jobs. The key attribute for the profile is category and value is
the category to which the job belongs to. For example you can set
the dagman category in the DAX for a job as follows
short-running-a top -T 6 -i -o
The property pegasus.dagman.[category].maxjobs can be used to
control the value.
For the above example, the user can set the property as follows
pegasus.dagman.short-running.maxjobs 2
In the DAG file generated you will see the category associated
with jobs. For the above example, it will look as follows
MAXJOBS short-running 2
CATEGORY preprocess_ID000001 short-running
JOB preprocess_ID000001 preprocess_ID000001.sub
RETRY preprocess_ID000001 2
20) Handling of pass through LFN
If a job in a DAX, specifies the same LFN as an input and an
output, it is a pass through LFN. Internally, the LFN is tagged
only as an input for the job. The reason for this, being that we
need to make sure that the replica catalog is queried for the
location of the LFN. If this is not handled specially, then LFN is
tagged internally as inout ( meaning it is generated during
workflow execution ). LFN's with type inout are not queried for in
the Replica Catalog in the force mode of operation
21) Tripping seqexec on first job failures
By default seqexec does not stop execution even if one of the
clustered jobs it is executing fails. This is because seqexec
tries to get as much work done as possible. If for some reason,
you want to make seqexec stop on first job failure, set the
following property in the properties file
pegasus.clusterer.job.aggregator.seqexec.firstjobfail true
22) New properties to choose the cleanup implementation
Two new properties were added to select the strategy and
implementation for file cleanup.
pegasus.file.cleanup.strategy
pegasus.file.cleanup.implementation
Currently there is only one cleanup strategy ( InPlace ) that can
be used and is loaded by default.
The cleanup implementations that can be used are
- Cleanup ( default)
- RM
- S3
Detailed documentation can be found at
$PEGASUS_HOME/etc/sample.properties.
23) New properties to choose the create dir implementation
The property pegasus.dir.create was deprecated.
It has been replaced by
pegasus.dir.create.strategy
Additionally, a user can specify a property to choose the
implementation used to create the directory on the remote sites.
pegasus.dir.create.impl
The create directory implementation that can be used are
- DefaultImplementation uses $PEGASUS_HOME/bin/dirmanager
executable to create a directory on the
remote site.
- S3 usese s3cmd to create a bucket on amazon S3.
BUGS FIXED
----------
1) Makefile for kickstart to build on Cygwin
Kickstart could not find SYS_NMLN variable in Cygwin to determine
the uname datastructure's size. Added a fix in the Makefile to add
CFLAGS -DSYS_NMLN=20 when the OS is Cygwin/Windows
2) Bug fix to getsystem release tools
Some systems have started using / in their system version name
which causes failures in Pegasus build process. Fixed the getsystem
release script which converts / into _
3) Bug fix in file cleanup module when stageout is enabled.
There was a bug in how the dependencies are added between the
stageout jobs and the file cleanup jobs. In certain cases, cleanup
could occur before the output was staged out. This is fixed now.
This bug was tracked through bugzilla
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=37
4) Bug fix to deferred planning
Deferred planning used to fail if pegasus-plan was not given -o
option .
This is fixed now and was tracked through bugzilla
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=34
5) Bug fix to caching on entries from Transformation Catalog
In certain cases, caching of entries did not work for the INSTALLED
case.
This is fixed now and was tracked through bugzilla
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=33
===============================
Release Notes for PEGASUS 2.1.0
===============================
NEW FEATURES
--------------
1) Support for Second Level Staging
Normally, Pegasus transfers the data to and from a directory on the
shared filesystem on the head node of a compute site. The directory
needs to be visible to both the head node and the worker nodes for
the compute jobs to execute correctly.
In the case, where the worker nodes cannot see the filesystem of
the head node there needs to be a Second Level Staging (SLS)
process that transfers the data from the head node to a directory
on the worker node tmp. To achieve this, Pegasus uses the pre-job
and post-job feature of kickstart to pull the input data from the
head node and push back the output data of a job to the head
node.
Even though we do SLS, Pegasus still relies on the existence of a
shared file system due to the following two reasons
a) for the transfer executable to pick up the proxy, that we
transfer from the submit host to the head node.
b) to access sls input and output files that contain the file
transfer urls to manage the transfer of data to worker node and
back to headnode.
Additionally, if you are running your workflows on a Condor pool,
one can bypass the use of kickstart to do the SLS. Please contact
pegasus@isi.edu for more details of this scenario. In this case,
the workflows generated by Pegasus have been shown to run in total
non shared filesystem environment.
To use this feature, user needs to set
pegasus.execute.*.filesystem.local true
The above change was tracked via bugzilla
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=21
2) New DAX schema
The new has release moved to the new DAX schema version 2.1. Schema
is available online http://pegasus.isi.edu/schema/dax-2.1.xsd
The main change in it is that the dontTransfer and dontRegister
flags have been replaced by transfer and register flags. Changes
were made both to the Java DAX Generator and Pegasus to conform to
the new schema.
Additionally, the DAX parser in Pegasus looks at the schema version to
determine whether to pick up dontTransfer and dontRegister flags (
to support backward compatibility with the older daxes).
Also with the filename type added a type attribute. It defaults to
data. Additionally user can have the values
executable|pattern. Users can use type=executable to specify any
dependant executables that their jobs required. All executable
files are tracked in the transformation catalog.
The above change was tracked via bugzilla
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=6
3) Workflow and Planner Metrics Logging
Workflow and Planning metrics are now logged for each workflow that
is planned by Pegasus. By default, they are logged to
$PEGASUS_HOME/var/pegasus.log
To turn metrics logging off, set pegasus.log.metrics to false
To change the file to which the metrics are logged set
pegasus.log.metrics.file path/to/log/file
Here is a snippet from the log file that shows what is logged
{
user = vahi
vogroup = pegasus-ligo
submitdir.base = /nfs/asd2/vahi/jbproject/Pegasus/dags
submitdir.relative = /vahi/pegasus-ligo/blackdiamond/run0064
planning.start = 2007-09-24T18:14:23-07:00
planning.end = 2007-09-24T18:14:29-07:00
properties
=/nfs/asd2/vahi/jbproject/Pegasus/dags/vahi/pegasus-ligo/blackdiamond/run0064/pegasus.6766.properties
dax = /nfs/asd2/vahi/jbproject/Pegasus/blackdiamond_dax.xml
dax-label = blackdiamond
compute-jobs.count = 3
si-jobs.count = 1
so-jobs.count = 3
inter-jobs.count = 0
reg-jobs.count = 3
cleanup-jobs.count = 2
total-jobs.count = 14
}
4) Support for querying multiple replica catalogs
Pegasus now allows the users to query multiple replica catalogs at
the same time to discover the locations of input data sets.
For this a new Replica Catalog implmentation was developed.
The users need to do the following to use it.
Set the replica catalog to MRC in the properties file.
pegasus.catalog.replica MRC
Each associated replica catalog can be configured via properties as
follows. The user associates a variable name referred to as [value]
for each of the catalogs, where [value] is any legal identifier
(concretely [A-Za-z][_A-Za-z0-9]*)
For each associated replica catalogs the user needs to specify the following properties.
pegasus.catalog.replica.mrc.[value] to specify the type of replica catalog
pegasus.catalog.replica.mrc.[value].key to specify a property name key for a
particular catalog
For example, if a user wants to query two lrc's at the same time he/she can specify as follows
pegasus.catalog.replica.mrc.lrc1 LRC
pegasus.catalog.replica.mrc.lrc2.url rls://sukhna
pegasus.catalog.replica.mrc.lrc2 LRC
pegasus.catalog.replica.mrc.lrc2.url rls://smarty
In the above example, lrc1, lrc2 are any valid identifier names and
url is the property key that needed to be specified.
5) Local Replica Selector
Pegasus has a new local replica selector that only prefers replicas
from the local host and that start with a file: URL scheme. It is
useful, when users want to stagin files to a remote site from your
submit host using the Condor file transfer mechanism.
In order to use this, set the replica selector to Local in the
properties.
- pegasus.selector.replica Local
6) Heft Based Site Selector
Added a new site selector that is based on the HEFT processor
scheduling algorithm.
The implementation assumes default data communication costs when jobs
are not scheduled on to the same site. Later on this may be made more
configurable.
The runtime for the jobs is specified in the transformation catalog
by associating the pegasus profile key runtime with the entries.
The number of processors in a site is picked up from the attribute
idle-nodes associated with the vanilla jobmanager of the site in
the site catalog.
To use this site selector, users need to set the following property
pegasus.selector.site Heft
7) Using multiple grid ftp servers for stageout
If a user specifies multiple grid ftp servers for the output site
in the site catalog, the stageout jobs will be distributed over all
of them.
More info can be found at
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=3
8) Scalable Directory structure on the stageout site
Users can now distribute their output files in a directory
structure on the output site. On setting the Boolean property
pegasus.dir.storage.deep to true, the relative submit directory
structure is replicated on the output site. Additionally, within
this directory the files are distributed into sub directories with
each subdirectory having 256 files.
The subdirectories are named in decimal format.
9) Specifying the jobmanager universe for the compute jobs in the DAX
Users can know specify the jobmanager type for the compute jobs in the
DAX. This is achieved by specifying the jobmanager.universe profile
key in the hints namespace.
Valid values for this are transfer|vanilla.
This is useful for users who are running on a grid site, with the
worker nodes behind a firewall and want a subset of their jobs to
run on the head node.
More info can be found at
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=5
10) Stork Support for doing transfers
The internal transfer interfaces of Pegasus were updated to use the
latest version of Stork for managing data transfers.
To use Stork implementations set the following
pegasus.transfer.refiner = SDefault
pegasus.transfer.*.impl=Stork
11) nogrid option for pegasus-run
pegasus-run has now a --nogrid option. This bypasses the checks for
proxy existence that are done before submitting the workflow for
execution. It disables all globus checks like check for environment
variables GLOBUS_lOCATION and LD_LIBRARY_PATH.
This is useful for running workflows in native Condor environments.
12) Submitting workflows directly using pegauss-plan
A new option --submit|-S option was added to pegasus-plan. This
allows users to submit workflows directly, after they have been
planned.
13) Specifying relative submit directory
Since pegasus 2.0 , pegasus-plan creates a directory structure in
the base submit directory. The base submit directory is specified
by --dir option to pegasus-plan. If a user, want to specify a
relative submit directory, he can use the --relative-dir option to
pegasus-plan.
The above change was tracked via bugzilla
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=14
BUGS FIXED
----------
1) Specifying Relative Path to the DAX
An incorrect path to the dax was generated internally when a user
specified a relative path to the dax to pegasus-plan
This is fixed now, and was tracked via bugzilla
http://vtcpc.isi.edu/bugzilla/show_bug.cgi?id=13
2) RLS java api bug fix 4114 (globus bugzilla number )
globus_rls_client.jar updated with the bug fix 4114. Also added the
jar for java 1.4 in lib/java1.4
3) Passing of DAGMan parameters via properties in case of deferred
planning
In case of deferred planning, the properties that control DAGMan
execution were not being passed as options to DAGMan. This is now
fixed.
The following properties are being handled correctly now.
pegasus.dagman.maxjobs
pegasus.dagman.maxpre
pegasus.dagman.maxidle
pegasus.dagman.maxpost
===============================
Release Notes for PEGASUS 2.0.1
================================
There is new documentation in the form of a quick start guide and glossary
in the docs directory. More documentation will be coming soon and will be available
in the release as well as on the pegasus website under documentation.
NEW FEATURES
------------
1) Pegasus now can store provenance data into PASOA. The actions taken
by the various refiners are logged into the store. It is still an
experimental feature. To turn it on, set the property
pegasus.catalog.provenance.refinement pasoa
The PASOA store needs to run on localhost on port 8080
https://localhost:8080/preserv-1.0
2) You can also use Pegasus to store execution provenance in PASOA.
To use set the properties
pegasus.exitcode.impl=pasoa
pegasus.exitcode.path.pasoa=${pegasus.home}/bin/pasoa-client
pegasus.exitcode.arguments=
BUGS FIXED
----------
1) sitecatalog-converter
patch to fix pegasus profile conversion
2) pegasus-submit-dag
added --maxidle option to allow setting number of idle jobs on
the remote site.
3) VORS.pm
Fixed a small typo in there, that lead to perl compilation
errors.
4) pegasus-get-sites
Removed local tc entries and added environments
for PEGASUS_HOME, GLOBUS_LOCATION and LD_LIBRARY_PATH to local
site.
5) mpiexec
The execution of clustered jobs via mpiexec was broken in 2.0
release. That is now fixed.
6) exitcode/exitpost
Fixed a bug in exitcode that caused a call to the DB PTC even though
the property pegasus.catalog.provenance was not set.
KNOWN BUGS
-----------
===============================
Release Notes for PEGASUS 2.0.0
===============================
NEW FEATURES
--------------
pegasus-plan.
This is the main client for invoking pegasus. The earlier gencdag command is now called pegasus-plan
pegasus-run
This is the client that submits the planned workflow to Condor and starts a monitoring tailstatd daemon
pegasus-status
This client lets you monitor a particular workflow. Its a wrapper around condor-q
pegasus-remove.
This client lets you remove a running workflow from the condor queue.
A rescue dag will be generated which can be submitted by just running pegasus-run on the dag directory.