Software Links
Getting Started
- Doc Structure
- A Globus Primer
- Globus Is Modular!
- Quickstart
- Installing GT
- Platform Notes
- Migrating from GT2
- Migrating from GT3
Reference
- PDF version
- Best Practices
- Coding Guidelines
- API docs
- Public Interfaces
- Resource Properties
- Samples
- Glossary
- Performance Studies
Common Runtime
Security
Data Mgt
Information Svcs
Execution Mgt
Table of Contents
RFT Service implementation in GT 4.1.1 uses standard SOAP messages over HTTP to submit and manage a set of 3rd party GridFTP transfers and to delete files using GridFTP. The user creates a RFT resource by submitting a list of URL pairs of files that need to be transferred/deleted to RFT Factory service. The user also specifies the time to live for the resource the user is creating to a GT 4.1.1 container in which RFT is deployed and configured. The resource is created after the user is properly authorized and authenticated. RFT service implementation exposes operations to control and manage the transfers (the resource). The operations exposed by both RFT factory and RFT service are briefly described below. The resource the user created also exposes the state of the transfer as a resource property to which the user can either subscribe for changes or poll for the changes in state periodically using standard command line clients.
Features new in GT 4.1.1
- No new features since GT 3.9.5.
Supported Features
- Delete files: Delete a set of files/directories on a GridFTP server.
- Exponential Backoff: Configurable exponential back off before a failed transfer is retried.
- Transfer All or None: If this option is set and one of the transfers in the request fails, RFT will stop transferring the remainder of the request and delete the files that were already transferred successfully.
- Transfer Permissions: File permissions are restored at the destination once the file is transfered successfully. This can be configured to throw a fatal error or a transient error depending on whether the GridFTP server supports the MLST command.
- Configurable number of concurrent transfers per container and per request.
- Better error reporting and faults.
- Database purge of the request and transfers after life time expiration.
- Cumulative (aggregate ) Resource Properties on the factory provide some statistical information.
- One status Resource Property for the entire transfer.
- Recursive directory transfers and deletes.
- Parallel streams.
- TCP Buffer Size.
- Third-party directory transfers, file transfers and deletes.
- Data channel authentication (DCAU).
- NoTPT option.
- Different subject names for source and destination GridFTP servers for the authorization mechanism.
- Support for binary/ascii type of transfers.
- Configurable number of retries for failed transfers per request.
- Block Size in bytes.
Deprecated Features
- None
Tested platforms for RFT:
Linux
- Fedora Core 1 i686
- Fedora Core 3 i686
- RedHat 7.3 i686
- RedHat 9 x86
- Debian Sarge x86
- Debian 3.1 i686
Mac OS X
- Mac OS X 10.3, 10.4
Tested containers for RFT:
- Java WS Core container
- Tomcat 5.0.30
Protocol changes since GT 4.0.4
- Added All or None option, maximum attempts, and finishBy to the transfer request
- Not backwards compatible with the OGSI version
API changes since GT 4.0.4
- None
Exception changes since GT 4.0.4
- None
Schema changes since GT 4.0.4
- WSDL changes to work with the new Java WS Core
RFT depends on the following GT components:
- Java WS Core
- WS Authentication and Authorization
- Delegation Service
- Service Groups
- MDS useful RP
RFT depends on the following 3rd party software:
- PostgreSQL 7.1 or later. Not tested with 8.0 yet.
The service configuration files such as jndi-config.xml and server-config.wsdd
(located under etc/<gar>/ directory) contain private information such as database passwords and usernames. Ensure that these configuration files are only readable by the user that is running the container.
The deployment process automatically sets the permissions of jndi-config.xml and server-config.wsdd as user readable only. However, this might not work correctly on all platforms and this does not apply to any other configuration files.
RFT stores the transfer requests in a database. Proper security measures need to be taken to protect the access of the data by granting/revoking appropriate permissions on tables that are created for RFT use and other steps that are appropriate and consistent with site specific security measures.
A design doc can be found here.
The Reliable Transfer Service (RFT) is a WSRF based service that
provides interfaces for controlling and monitoring third party file
transfers using GridFTP servers. The client controlling the transfers
(in this case RFT) is hosted inside of a Grid service so it can
be managed using the soft state model. It is essentially a reliable and
recoverable version of the GT2 globus-url-copy tool and more. In GT 4.1.1, RFT
can also perform file deletion and recursive directory deletion operations. It is also used by GRAM to perform all the staging
operations and cleanup operations.
The RFT service implementation in GT 4.1.1 uses standard SOAP messages over HTTP to submit and manage a set of 3rd party GridFTP transfers and to delete files using GridFTP. The user creates an RFT resource by submitting a list of URL pairs of files that need to be transferred/deleted to the RFT Factory service. The user also specifies the time to live for the resource the user is creating to the GT 4.1.1 container in which RFT is deployed and configured. The resource is created after the user is properly authorized and authenticated. RFT service implementation exposes operations to control and manages the transfers (the resource). The operations exposed by both the RFT factory and the RFT service are briefly described below. The resource the user created also exposes the state of the transfer as a resource property to which the user can either subscribe for changes or poll for the changes in state periodically using standard command line clients.
Please find below operations of both RFT Factory and RFT Service Implementation.
Used to create a Reliable File Transfer resource. The operations exposed by the factory are as follows:
createReliableFileTransfer: Creates a Reliable File Transfer resource.- Input Parameters: Initial Termination time, Transfer Request or Delete Request.
- Output parameters: Termination time, Current time, Endpoint reference of the Resource created. This should be stored by the user, as it is needed to query the status of the resource and to perform any further operations on the resource.
- Fault: createReliableFileTransferFault.
Used to manage the Resource created using the RFT Factory Service. The operations exposed by the service are as follows:
start:Starts executing the transfers/deletes.- Input Parameters: None
- Output Parameters: None
- Fault: RepeatedlyStartedFault
getStatus:To get the status of a particular file.- Input Parameters: A source URL of the file that is part of the request.
- Output Parameters:
Transfer Status Type - Fault: RFTDatabaseFault
getStatusSet: To get the status of a set of files in a request.- Input Parameters: int from (the relative position of the transfer in the request) and int offset (the number of files queried).
- Output Parameters: An array of
TransferStatusType. - Fault: RFTDatabaseFault
cancel: To cancel a transfer that is part of a resource.- Input Parameters: int from (the relative position of the transfer in the request) and int to.
- Output Parameters: None
- Fault: RFTDatabaseFault
The resource properties of RFT Factory (which acts both as a resource and a service at the same time) and RFT Resource are found below:
ActiveResourceInstances: A dynamic resource property of the total number of active RFT resources in the container at a given point of time.TotalNumberOfTransfers: A dynamic resource property of the total number of transfers/deletes performed since the RFT service was deployed in this container.TotalNumberOfActiveTransfers: A dynamic resource property of the number of active transfers across all rft resources in a container at a given point of time.TotalNumberOfBytesTransferred: A dynamic resource property of the total number of bytes transferred by all RFT resources created since the deployment of the service.RFTFactoryStartTime: Time when the service was deployed in the container. Used to calculate uptime.DelegationServiceEPR: The end point reference of the Delegation resource that holds the delegated credential used in executing the resource.
OverallStatus: This is a complex type providing the overall status of an RFT resource by providing the number of transfers pending, active, finished, retrying, failed, and cancelled. Each of these values can be obtained by invoking getTransfers(Finished/Active/Failed/Restarted/Pending/Cancelled) on OverallStatus Resource Property. Note that this Resource Property gets updated every time one of the transfers changes state, so there can be and will be more than one update in the life time of a RFT resource if you subscribe to this RP. This Resource Property also includes the last fault (if thrown) from a transfer and can be accessed by invoking getFault on OverallStatus. This will indicate why a transfer has failed.RequestStatus: This is a complex type resource property providing the status of an RFT resource in the form of Pending/Active/Done/Failed. The status can be obtained from RequestStatusType by invoking getRequestStatus(). This will result in one of four status strings (Pending/Active/Done/Failed/Cancelled). This RP also contains a fault that denotes the last fault in a RFT resource and can be accessed by invoking getFault(). If a client is subscribed to this RP, there will be only be 2 updates in the life time of an RFT resource (Pending->Active->Done, Pending->Active->Failed, Pending->Active->Cancelled, and Pending->Cancelled).TotalBytes: This provides the total number of bytes transferred by the resource.TotalTime: This provides the total time taken to transfer the above-mentioned total bytes.
Faults from the RFT Factory Service and RFT Service can be found below:
createReliableFileTransferFault: All the errors encountered during the creation of the RFT resource are mapped to this fault. Any security related errors are caught before the factory and are thrown to the user/client.
You can find links to all the RFT schemas here.
Please see the RFT Command Reference.
Please go here to view the entire RFT transfer request schema documentation.
These options are set in the transferRequest and deleteRequest elements and apply similarly for each.
concurrency
This denotes number of files in the request that needs to be transferred at one time.
maxAttempts
Maximum number of attempts after transient errors to execute the transfer or deletion before giving up and raising an error.
finishBy
(Not Implemented) In future versions of RFT this will be used to enforce time constraints on a transfer.
These options are set in the
rftOptions element
(see
RFTOptionsType
for more details)
and are specific
to file transfers. They can be specified as defaults for all transfers
under the
transferRequest
element,
and/or individually under each transfer
element
(see
TransferType
for more details):
<transferRequest>
<transfer>...</transfer>
<rftOptions>
<-- option elements here -->
</rftOptions>
</transferRequest>
AND/OR
<transferRequest>
<transfer>
...
<sourceUrl>
<destinationUrl>
...
<rftOptions>
<-- option elements here -->
</rftOptions>
</transfer>
</transferRequest>
binary
Transfer as a binary file. Default is "true".
blockSize
Specifies the size of the data blocks to use in the transfer.
tcpBufferSize
Specifies the TCP buffer size used for the transfer.
notpt
If set to "true", third-party transfer mode will not be use. Instead, a client thread will be started that will GET data from the source server and and PUT data to the destination server. Default is "false".
parallelStreams
Specifies the number of parallel streams to use during the transfer. Default is 1.
dcau
Specifies whether or not to use data channel authentication. Default is true.
subjectName
Specifies the credential subject to use for authenticating both the source and destination servers.
destinationSubjectName
Specifies the credential subject to use for authenticating the destination server.
sourceSubjectName
Specifies the credential subject to use for authenticating the source server.
userName
Specifies the username to be used to perform the transfer which sometimes may not be the same as transfer requester.
These options are set in the
deleteOptions element
(see
DeleteOptionsType
for more details),
and are specific to file deletions. They can be specified as defaults for
all deletions under the
deleteRequest
element, and/or individually under each
deletion
element
(see
DeleteType
for more details):
<deleteRequest>
<deletion>...</deletion>
<deleteOptions>
<-- option elements here -->
</deleteOptions>
</deleteRequest>
AND/OR
<deleteRequest>
<deletion>
...
<file>
<deleteOptions>
<-- option elements here -->
</deleteOptions>
</deletion>
</deleteRequest>
subjectName
Specifies the credential subject to use for authenticating the target server.
userName
Specifies the username to be used to perform the deletion.
RFT has the following prerequisites:
- Java WS Core - This is built and installed in a GT 4.1.1 System Administrator's Guide.
- A host certificate (see GT 4.1.1 System Administrator's Guide).
- GridFTP - GridFTP performs the actual file transfer and is built and installed in a GT 4.1.1 System Administrator's Guide.
- PostgreSQL - PostgreSQL is used to store the state of the transfer to allow for restart after failures. The interface to PostgreSQL is JDBC, so any DBMS that supports JDBC can be used, although no others have been tested. For instructions on configuring the PostgreSQL database for RFT, see below. .
The security of the service can be configured by modifying the security descriptor. It allows for configuring the credentials that will be used by the service, type of authentication and authorization that needs to be enforced. By default, the following security configuration is installed:
- Credentials set for use by the container are used. If they arenot specified, default credentials are used.
- GSI Secure conversation authentication is enforced for all methods.
Note: Changing the required authentication and authorization method will require suitable changes to the clients that contact this service.
To alter the security descriptor configuration, refer to security descriptor. The file to be altered is $GLOBUS_LOCATION/etc/globus_wsrf_rft/security-config.xml.
PostgreSQL (version 7.1 or greater) needs to be installed and configured for RFT to work. You can either use the packages which came with your operating system (RPMs, DEBs, ...) or build from source. We used PostgreSQL version 7.3.2 for our testing and the following instructions are good for the same.
Install PostgreSQL. Instructions on how to install/configure PostgreSQL can be found here.
Configure the postmaster daemon so that it accepts TCP connections. This can be done by adding the "-o -i" switch to the postmaster script (This is either the init.d script found in /etc/init.d/postgresql or /var/lib/, depending on how you installed PostgreSQL). Follow the instructions here to start the postmaster with the -i option.
You will now need to create a PostgreSQL user that will connect to the database. This is usually the account under which the container is running. You can create a PostgreSQL user by running the following command:
su postgres; createuser globus. If you get the following error:psql: could not connect to server: No such file or directory Is the server running locally and accepting connections on Unix domain socket "/tmp/.s.PGSQL.5432"?this generally means that either your postmaster is not started with the -i option or you didn't restart the postmaster after the above mentioned step.Now you need to set security on the database you are about to create. You can do it by following the steps below:
sudo vi /var/lib/pgsql/data/pg_hba.confand append the following line to the file:host rftDatabase "username" "host-ip" 255.255.255.255 md5Note: use crypt instead of md5 if you are using PostgreSQL 7.3 or earlier.sudo /etc/init.d/postgresql restartTo create the database that is used for RFT run (as user globus):
createdb rftDatabase.To populate the RFT database with the appropriate schemas run:
psql -d rftDatabase -f $GLOBUS_LOCATION/share/globus_wsrf_rft/rft_schema.sql. Now that you have created a database to store RFT's state, the following steps configure RFT to find the database:Open
$GLOBUS_LOCATION/etc/globus_wsrf_rft/jndi-config.xml.Find the
dbConfigurationsection under theReliableFileTransferService <service>section.Change the
connectionStringto point to the machine on which you installed PostgreSQL and to the name of the database you used in step 2. If you installed PostgreSQL on the same machine as your Globus install, the default should work fine for you.Change the
userNameto the name of the user who owns/created the database and do the same for the password (it also depends on how you configured your database).Don't worry about the other parameters in the section. The defaults should work fine for now.
Edit the configuration section under
ReliableFileTransferService. There are two values that can be edited in this section:backOff: Time in seconds you want RFT to backoff before a failed transfer is retried by RFT. The default should work fine for now.maxActiveAllowed: This is the number of transfers the container can do at given point. The default should be fine for now.
With a default GT 4.1.1 installation, the RFT service is automatically registered with the default WS MDS Index Service running in the same container for monitoring and discovery purposes.
There is a jndi resource defined in $GLOBUS_LOCATION/etc/globus_wsrf_rft/jndi-config.xml as follows :
<resource name="mdsConfiguration"
type="org.globus.wsrf.impl.servicegroup.client.MDSConfiguration">
<resourceParams>
<parameter>
<name>reg</name>
<value>true</value>
</parameter>
<parameter>
<name>factory</name>
<value>org.globus.wsrf.jndi.BeanFactory</value>
</parameter>
</resourceParams>
</resource>
To configure the automatic registration of RFT to the default WS MDS Index Service, change the value of the parameter
<reg> as follows:
trueturns on auto-registration; this is the default in GT 4.1.1.falseturns off auto-registration.
By default, the following resource properties (from the RFT Factory Resource) are sent to the default Index Service:
ActiveResourceInstances: A dynamic resource property of the total number of active RFT resources in the container at a given point of time.TotalNumberOfTransfers: A dynamic resource property of the total number of transfers/deletes performed since the RFT service was deployed in this container.TotalNumberOfActiveTransfers: A dynamic resource property of the number of active transfers across all rft resources in a container at a given point of time.TotalNumberOfBytesTransferred: A dynamic resource property of the total number of bytes transferred by all RFT resources created since the deployment of the service.RFTFactoryStartTime: Time when the service was deployed in the container. Used to calculate uptime.DelegationServiceEPR: The end point reference of the Delegation resource that holds the delegated credential used in executing the resource.
You can configure which resource properties are sent in RFT's registration.xml file, $GLOBUS_LOCATION/etc/globus_wsrf_rft/registration.xml.
The following is the relevant section of the file:
<Content xsi:type="agg:AggregatorContent"
xmlns:agg="http://mds.globus.org/aggregator/types">
<agg:AggregatorConfig xsi:type="agg:AggregatorConfig">
<agg:GetMultipleResourcePropertiesPollType
xmlns:rft="http://www.globus.org/namespaces/2004/10/rft">
<!-- Specifies that the index should refresh information
every 60000 milliseconds (once per minute) -->
<agg:PollIntervalMillis>60000</agg:PollIntervalMillis>
<!-- specifies that all Resource Properties should be
collected from the RFT factory -->
<agg:ResourcePropertyNames>rft:TotalNumberOfBytesTransferred</agg:ResourcePropertyNames>
<agg:ResourcePropertyNames>rft:TotalNumberOfActiveTransfers</agg:ResourcePropertyNames>
<agg:ResourcePropertyNames>rft:RFTFactoryStartTime</agg:ResourcePropertyNames>
<agg:ResourcePropertyNames>rft:ActiveResourceInstances</agg:ResourcePropertyNames>
<agg:ResourcePropertyNames>rft:TotalNumberOfTransfers</agg:ResourcePropertyNames>
</agg:GetMultipleResourcePropertiesPollType>
</agg:AggregatorConfig>
<agg:AggregatorData/>
</Content>
If a third party needs to register an RFT service manually, see Registering with mds-servicegroup-add in the WS MDS Aggregator Framework documentation.
RFT is primarily used to reliably transfer large datasets using GridFTP. If you are a developer and would like to use RFT, the following steps would help you to do that.
Contact the Delegation Factory Service and get an EPR for the Delegation Resource that contains your delegated credential.
public static EndpointReferenceType delegateCredential(String host, String port) throws Exception { ClientSecurityDescriptor desc = new ClientSecurityDescriptor(); // Credential to sign with, assuming default credential GlobusCredential credential = GlobusCredential.getDefaultCredential(); desc.setGSITransport(Constants.GSI_TRANSPORT) Util.registerTransport(); desc.setAuthz('host'); String factoryUrl = PROTOCOL + "://" + host + ":" + port + SERVICE_URL_ROOT + DelegationConstants.FACTORY_PATH; // lifetime in seconds int lifetime = TERM_TIME * 60; // Get the public key to delegate on. EndpointReferenceType delegEpr = AddressingUtils .createEndpointReference(factoryUrl, null); X509Certificate[] certsToDelegateOn = DelegationUtil .getCertificateChainRP(delegEpr, desc); X509Certificate certToSign = certsToDelegateOn[0]; return DelegationUtil.delegate(factoryUrl, credential, certToSign, lifetime, false, desc); }Now construct a TransferRequestType Object:
TransferType[] transferArray = new TransferType[1]; transferArray[0] = new TransferType(); transferArray[0].setSourceUrl("gsiftp://foo/bar"); transferArray[0].setDestinationUrl("gsiftp://blah/"); RFTOptionsType rftOptions = new RFTOptionsType(); rftOptions.setBinary(true); // You can set more options like parallel streams, buffer sizes etc // Refer to Public Interface guide of RFT for more details TransferRequestType request = new TransferRequestType(); request.setRftOptions(rftOptions); request.setTransfer(transferArray); request.setTransferCredentialEndpoint(delegateCredential(host,port));Now contact the RFT factory and create an RFT resource:
public static EndpointReferenceType createRFT(String rftFactoryAddress, BaseRequestType request) throws Exception { endpoint = new URL(rftFactoryAddress); factoryPort = rftFactoryLocator .getReliableFileTransferFactoryPortTypePort(endpoint); CreateReliableFileTransferInputType input = new CreateReliableFileTransferInputType(); //input.setTransferJob(transferType); if(request instanceof TransferRequestType) { input.setTransferRequest((TransferRequestType)request); } else { input.setDeleteRequest((DeleteRequestType)request); } Calendar termTime = Calendar.getInstance(); termTime.add(Calendar.HOUR, 1); input.setInitialTerminationTime(termTime); setSecurity((Stub)factoryPort); CreateReliableFileTransferOutputType response = factoryPort .createReliableFileTransfer(input); return response.getReliableTransferEPR(); }Now contact the RFT service Implementation and call start to actually start the transfer:
ReliableFileTransferPortType rft = rftLocator .getReliableFileTransferPortTypePort(rftepr); setSecurity((Stub)rft); //For secure notifications subscribe(rft); System.out.println("Subscribed for overall status"); //End subscription code Calendar termTime = Calendar.getInstance(); termTime.add(Calendar.MINUTE, TERM_TIME); SetTerminationTime reqTermTime = new SetTerminationTime(); reqTermTime.setRequestedTerminationTime(termTime); System.out.println("Termination time to set: " + TERM_TIME + " minutes"); SetTerminationTimeResponse termRes = rft .setTerminationTime(reqTermTime); StartOutputType startresp = rft.start(new Start());
RFT can also be used to delete a set of files and directories using GridFTP server. The following steps depict how to:
Contact the Delegation Factory Service and get an EPR for the Delegation Resource that contains your delegated credential.
public static EndpointReferenceType delegateCredential(String host, String port) throws Exception { ClientSecurityDescriptor desc = new ClientSecurityDescriptor(); // Credential to sign with, assuming default credential GlobusCredential credential = GlobusCredential.getDefaultCredential(); desc.setGSITransport(Constants.GSI_TRANSPORT) Util.registerTransport(); desc.setAuthz('host'); String factoryUrl = PROTOCOL + "://" + host + ":" + port + SERVICE_URL_ROOT + DelegationConstants.FACTORY_PATH; // lifetime in seconds int lifetime = TERM_TIME * 60; // Get the public key to delegate on. EndpointReferenceType delegEpr = AddressingUtils .createEndpointReference(factoryUrl, null); X509Certificate[] certsToDelegateOn = DelegationUtil .getCertificateChainRP(delegEpr, desc); X509Certificate certToSign = certsToDelegateOn[0]; return DelegationUtil.delegate(factoryUrl, credential, certToSign, lifetime, false, desc); }Now construct a DeleteRequestType object:
DeleteType[] deleteArray = new DeleteType[1]; deleteArray[0] = new DeleteType(); deleteArray[0].setFile("gsiftp://foo/bar"); DeleteOptionsType deleteOptions = new DeleteOptionsType(); deleteOptions.setSubjectName("SUBJECT-NAME"); DeleteRequestType request = new DeleteRequestType(); request.setDeleteOptions(deleteOptions); request.setDeletion(deleteArray); request.setTransferCredentialEndpoint(delegateCredential(host,port));Now contact the RFT factory and create an RFT resource:
public static EndpointReferenceType createRFT(String rftFactoryAddress, BaseRequestType request) throws Exception { endpoint = new URL(rftFactoryAddress); factoryPort = rftFactoryLocator .getReliableFileTransferFactoryPortTypePort(endpoint); CreateReliableFileTransferInputType input = new CreateReliableFileTransferInputType(); //input.setTransferJob(transferType); if(request instanceof TransferRequestType) { input.setTransferRequest((TransferRequestType)request); } else { input.setDeleteRequest((DeleteRequestType)request); } Calendar termTime = Calendar.getInstance(); termTime.add(Calendar.HOUR, 1); input.setInitialTerminationTime(termTime); setSecurity((Stub)factoryPort); CreateReliableFileTransferOutputType response = factoryPort .createReliableFileTransfer(input); return response.getReliableTransferEPR(); }Now contact the RFT service Implementation and call start to actually start the transfer:
ReliableFileTransferPortType rft = rftLocator .getReliableFileTransferPortTypePort(rftepr); setSecurity((Stub)rft); //For secure notifications subscribe(rft); System.out.println("Subscribed for overall status"); //End subscription code Calendar termTime = Calendar.getInstance(); termTime.add(Calendar.MINUTE, TERM_TIME); SetTerminationTime reqTermTime = new SetTerminationTime(); reqTermTime.setRequestedTerminationTime(termTime); System.out.println("Termination time to set: " + TERM_TIME + " minutes"); SetTerminationTimeResponse termRes = rft .setTerminationTime(reqTermTime); StartOutputType startresp = rft.start(new Start());
A standard way to debug RFT is to make the container print out more verbose error messages. You can do this with the following steps:
Edit $GLOBUS_LOCATION/container-log4j.properties and add following line to it:
log4j.category.org.globus.transfer=DEBUG. For more verbosity add log4j.category.org.globus.ftp=DEBUG , which will print out Gridftp messages too.
Database configuration is the most complicated and important part of RFT setup. You can find more instructions on troubleshooting in Section 8, “Troubleshooting”.
RFT uses PostgreSQL to check-point transfer state in the form of restart markers and recover from transient transfer failures, using retry mechanism with exponential backoff, during a transfer. RFT has been tested to recover from source and/or destination server crashes during a transfer, network failures, container failures (when the machine running the container goes down), file system failures, etc. RFT Resource is implemented as a PersistentResource, so ReliableFileTransferHome gets initialized every time a container gets restarted. Please find a more detailed description of fault-tolerance and recovery in RFT below:
- Source and/or destination GridFTP failures: In this case RFT retries the transfer for a configurable number of maximum attempts with exponential backoff for each retry (the backoff time period is configurable also). If a failure happens in the midst of a transfer, RFT uses the last restart marker that is stored in the database for that transfer and uses it to resume the transfer from the point where it failed, instead of restarting the whole file. This failure is treated as a container-wide backoff for the server in question. What this means is that all other transfers going to/from that server, across all the requests in a container, will be backed off and retried. This is done in order to prevent further failures of the transfers by using knowledge available in the database.
- Network failures: Sometimes this happens due to heavy load on a network or for any other reason packets are lost or connections get timed out. This failure is considered a transient failure and RFT retries the transfer with exponential backoff for that particular transfer (and not the whole container, as with the source and/or destination GridFTP failures).
- Container failures: These type of failures occur when the machine running the container goes down or if the container is restarted with active transfers. When the container is restarted, it restarts ReliableTransferHome, which looks at the database for any active RFT resources and restarts them.