Changes for page User Guide: Swift Transfer with FTS
Last modified by luehrs on 2023/04/12 09:26
Summary
-
Page properties (1 modified, 0 added, 0 removed)
-
Attachments (0 modified, 0 added, 2 removed)
Details
- Page properties
-
- Content
-
... ... @@ -16,204 +16,4 @@ 16 16 17 17 [[image:image-20230412090045-1.png]] 18 18 19 -== Requirements == 20 - 21 -To use the service correctly, the user needs 22 - 23 -* valid credentials (to access both to source hosting site and destination hosting site); 24 -* valid account(s) (defined both at the source hosting site and at the destination hosting site); 25 -* at least one object storage as the source; and 26 -* at least one object storage as the destination. 27 - 28 -Additionally, the user needs: 29 - 30 -* the FTS REST client; and 31 -* a valid token to be mapped. 32 - 33 -In case the user would like to check if something is missing, she/he can open an issue with the respective local User Support services of ICEI hosting sites. 34 - 35 -**Object store URLs** 36 - 37 -Presently, only three ICEI sites [[are providing object storages to the HBP community>>https://fenix-ri.eu/infrastructure/resources/available-resources]]: BSC (ES), CSCS (CH), and JSC (DE). As soon as other sites will start the provisioning of these resources, the list of URLs below will be updated. 38 - 39 -* CSCS: [[https:~~/~~/object.cscs.ch>>https://object.cscs.ch]] 40 -* BSC: [[https:~~/~~/swift.bsc.es>>https://swift.bsc.es]] 41 -* JSC: [[https:~~/~~/just-object.fz-juelich.de:8080>>https://just-object.fz-juelich.de:8080]] 42 - 43 -Presently the Central Data Transfer service, to transfer data between object storages, is provided at BSC. The endpoint of the service is reported below. In case other ICEI sites will host the service, the list below will be updated. 44 - 45 -**FTS endpoint**: [[https:~~/~~/fts.bsc.es:8446>>https://fts.bsc.es:8446]] 46 - 47 -== How to obtain... == 48 - 49 -In this section, the user will find some useful instructions to solve the requirements. 50 - 51 -=== Valid credentials === 52 - 53 -To access the service, the user needs to be defined in the local user databases, and in FURMS user database. Valid credentials (username and password) are provided by ICEI local sites, according to Fenix and (possible) local policies applied, if the user is a Principal Investigator or Collaborator in an approved project. Please note that proposal submission to obtain resources is open every working day for the HBP community. More information is available at: [[https:~~/~~/wiki.ebrains.eu/bin/view/Collabs/fenix-icei/>>url:https://wiki.ebrains.eu/bin/view/Collabs/fenix-icei/]] 54 - 55 -=== Valid account(s) and object storages === 56 - 57 -To obtain a valid account on FURMS, the user needs to register at [[https:~~/~~/ebrains.eu/register/>>url:https://ebrains.eu/register/]]. After her/his proposal (sub-section above) has been approved, and the project has been defined at the hosting sites, the user will receive valid local credentials, too. 58 - 59 -To obtain object storage resources, the user needs to request them explicitly in the submission form of the proposal, asking for ARD resources and specifying that object storage resources are needed. Please note that the submission form allows applicants to ask for resources distributed across ICEI sites with a single proposal. 60 - 61 -=== FTS REST client === 62 - 63 -To perform a data transfer through the FTS server, you need to install the fts-rest-client. The RPM package will provide the CLI tools for the FTS server. The tested version is available in the EBRAINS collaboratory bucket: [[https:~~/~~/data-proxy.ebrains.eu/api/v1/permalinks/4780cb34-8c71-497e-85bd-3bf58a9fc21e>>url:https://data-proxy.ebrains.eu/api/v1/permalinks/4780cb34-8c71-497e-85bd-3bf58a9fc21e]]. 64 - 65 -To install the RPM package one needs to run the following command: 66 - 67 -##[promt]$ yum install fts-rest-client-1.0.0-%{_release}.el7.noarch## 68 - 69 -Should the user prefer to build the fts-rest-client him/herself, the scripts can be found in the fts-rest-flask repository here: [[https:~~/~~/gitlab.cern.ch/fts/fts-rest-flask/-/tree/swift>>url:https://gitlab.cern.ch/fts/fts-rest-flask/-/tree/swift]]. The user can clone the repository, switch to the swift branch, and ##cd## into it. We advise creating a ##venv## for this project. Then run ##pip install .## or ##python setup.py## install to install the fts package. After that, the user may run the scripts in src/cli the same way he/she uses the FTS commands. 70 - 71 -=== Valid OIDC token === 72 - 73 -To obtain a valid OIDC token, the user needs to login via her/his institute’s IdP or HBP user account, and she/he should be able to see her/his access token on the [[https:~~/~~/fts.bsc.es>>url:https://fts.bsc.es]] website. The user should copy the token to run the FTS commands. We advise all users to perform this action at least once to have their credentials registered with FTS properly. Please, note that this is a temporary solution for retrieving the access token. 74 - 75 -For CLI users, Device Code Flow ([[RFC 8628>>url:https://www.rfc-editor.org/rfc/rfc8628.html]]) is the recommended authorisation grant to obtain OIDC tokens. The user can install [[oidc-agent>>url:https://indigo-dc.gitbook.io/oidc-agent/]] to perform such workflows. To use oidc-agent with Fenix AAI, a development version 4.4.1 is needed; such packages can be found here: [[https:~~/~~/repo.data.kit.edu/devel/>>url:https://repo.data.kit.edu/devel/]]. 76 - 77 -1. Add the following line to /etc/oidc-agent/pubclients.config (for the first time): 78 -##APP-EF25A717-2059-461D-8275-A71CC098B82B@https:~/~/central-proxy.fenix-ri.eu## 79 -1. Initiate oidc-agent: 80 -##[promt]$ eval $(oidc-agent)## 81 -1. The user needs to run the oidc-gen command to generate account info for Fenix: 82 -##[promt]$ oidc-gen ~-~-pub ~-~-issuer https:~/~/central-proxy.fenix-ri.eu \ 83 -~-~-configuration-endpoint https:~/~/central-webapp.fenix-ri.eu/oidc/.well-known/openid-configuration \ 84 -~-~-flow device ~-~-no-save <define_an_account_name>## 85 -Follow the instructions of the output, login to Fenix AAI. 86 -1. Obtain an OIDC token: 87 -##[promt]$ oidc-token <defined_account_name>## 88 - 89 -== How to perform a data transfer with FTS == 90 - 91 -In this section, the user can find useful commands and examples. More detailed commands are available at [[FTS3 Documentation - CERN>>url:https://fts3-docs.web.cern.ch/fts3-docs/]]. 92 - 93 -=== FTS commands === 94 - 95 -The FTS transfers are managed by the following commands: 96 - 97 -* fts-rest-transfer-submit //to submit transfer job// 98 -* fts-rest-transfer-status //to query a transfer job// 99 -* fts-rest-transfer-list //to list transfer jobs// 100 -* fts-rest-transfer-cancel //to cancel a transfer job// 101 - 102 -=== Source/destination URL format === 103 - 104 -FTS is a protocol-specific transfer service. If the user wants to transfer from/to a Swift endpoint, she/he should specify ##swifts:~/~/ + storage name## instead of ##https:~/~/ + storage name## in the source/destination URL. The ##swifts:~/~/## is to specify she/he is using Swift protocol with FTS. 105 - 106 -The format of a Swift url to submit to FTS is ##swifts:~/~/ + storage name + container name + (folder name(s)) + object name##. 107 - 108 -For example, ##swifts:~/~/swift.bsc.es/testcontainer/data.txt## locates the data.txt file in container testcontainer in the BSC object store. 109 - 110 -=== Submit a Swift-to-Swift transfer === 111 - 112 -We give instructions in this manual for Swift-to-Swift data transfer only, because the current available Fenix archival repositories are Swift object stores. For other types of storage, users may refer to the official FTS guide [[FTS3 Documentation - CERN>>url:https://fts3-docs.web.cern.ch/fts3-docs/]]. 113 - 114 -For Swift-to-Swift transfer, the user has to specify OS project ids which indicate the data source and destinations. 115 - 116 -Users can find the project id by either: 117 - 118 -1. using the command ##openstack project list## with the Openstack CLI client. 119 -1. login to the Openstack dashboard, go to //Identity -> Projects//. 120 -For example, on [[https:~~/~~/ncloud.bsc.es>>url:https://ncloud.bsc.es]]: 121 -[[image:image-20230412091641-1.png]] 122 -and on [[https:~~/~~/castor.cscs.ch>>url:https://castor.cscs.ch]]: 123 -[[image:image-20230412091702-2.png]] 124 - 125 -To transfer from/to the object storages at JSC or CSCS, the user also needs to specify an OS token at the moment. 126 - 127 -* For **JSC** users, please refer to the documentation [[here>>url:https://apps.fz-juelich.de/jsc/hps/just/object-storage.html#user-environment-preparation]] to create the ##rc## file, then do ##openstack token issue##, your OS token will be shown in the “id” field. 128 -* For **CSCS** users, the easiest way to get an OS token is to follow the guide on this GitHub page: [[https:~~/~~/github.com/eth-cscs/openstack/tree/master/cli>>url:https://github.com/eth-cscs/openstack/tree/master/cli]]. When authenticated, the OS token is saved as environment variable ##OS_TOKEN##. 129 -* If the user would like to specify the OS token for **BSC** her/himself (which is not necessary), guidance is provided here: [[https:~~/~~/bsc.es/supportkc/docs-ncloud/access#cli-access>>url:https://bsc.es/supportkc/docs-ncloud/access#cli-access]]. 130 - 131 -Below is an example of submitting the transfer with CLI: 132 - 133 -##$fts-rest-transfer-submit \ 134 --s https:~/~/fts.bsc.es:8446/ \## 135 -##~-~-access-token $tok \## 136 -##~-~-os-project-id “<source_project_id>:<dest_project_id>” \## 137 -##~-~-os-token “<project_id>:<corresponding_os_token>” \## 138 -##swifts:~/~/object.cscs.ch/<container_name>/<object_name> \## 139 -##swifts:~/~/swift.bsc.es/<container_name>/<object_name>## 140 - 141 -### Using endpoint: https:~/~/fts.bsc.es:8446 142 -# REST API version: 3.10.0 ## 143 -##Job successfully submitted.## 144 -##Job id: 532086a8-30a9-11eb-b0da-fa163e1833c1## 145 - 146 -Users can use multiple ##~-~-os-token## to specify OS tokens for different projects. 147 - 148 -==== Bulk transfer ==== 149 - 150 -Users can use bulk transfer by using the ##~-~-file## option. Then they have to submit a JSON file. 151 - 152 -##$fts-rest-transfer-submit \ 153 --s https:~/~/fts.bsc.es:8446/ \## 154 -##~-~-access-token $tok \## 155 -##~-~-file <your_file>.json## 156 - 157 -The format of such files is described here: [[https:~~/~~/fts3-docs.web.cern.ch/fts3-docs/fts-rest/docs/bulk.html>>url:https://fts3-docs.web.cern.ch/fts3-docs/fts-rest/docs/bulk.html]]. Note that the files the user transfers this way still need to be from and to the same object store, i.e., she/he can only transfer a list of files from project A at site 1 to project B at site 2 in one job. 158 - 159 -As it is the same for any Swift-to-Swift transfer, users have to supply OS project ids, and in some cases, OS tokens as well. Users can use the command line options to do so or put it in the ##params## section when using a JSON file. An example of the JSON file for a bulk transfer is shown below: 160 - 161 -##{ 162 -"files": [ 163 -{ 164 - "sources": [ 165 - "swifts:~/~/swift.bsc.es/testcontainer/cp1" 166 - ], 167 - "destinations": [ 168 - "swifts:~/~/object.cscs.ch/testcontainer/bulk/cp1" 169 - ] 170 -}, 171 -{ 172 - "sources": [ 173 - "swifts:~/~/swift.bsc.es/testcontainer/cp2" 174 - ], 175 - "destinations": [ 176 - "swifts:~/~/object.cscs.ch/testcontainer/bulk/cp2" 177 - ] 178 -} 179 -], 180 -"params":{ "os_project_id":"c7ccbe370791496fa44632e6ccb10405:7cc304cc62d44ab180d4a28427e2b64b", 181 - "os_token":[ 182 - "7cc304cc62d44ab180d4a28427e2b64b:gAAAAA…" 183 - ] 184 -}## 185 - 186 -=== Query a transfer === 187 - 188 -This command checks the status of the transfer job; a detailed FTS job state machine can be found here: [[https:~~/~~/fts3-docs.web.cern.ch/fts3-docs/docs/state_machine.html>>url:https://fts3-docs.web.cern.ch/fts3-docs/docs/state_machine.html]]. 189 - 190 -##$fts-rest-transfer-status \ 191 --s https:~/~/fts.bsc.es:8446/ \ 192 -~-~-access-token $tok \ 193 -<your_job_id>## 194 - 195 -The job ID is given when the user submits the transfer. 196 - 197 -=== List transfers === 198 - 199 -This command lists the ongoing transfers at the FTS endpoint. 200 - 201 -##$fts-rest-transfer-list \ 202 --s https:~/~/fts.bsc.es:8446/ \ 203 -~-~-access-token $tok## 204 - 205 -=== Cancel a transfer === 206 - 207 -This command cancels the transfer job. 208 - 209 -##$fts-rest-transfer-cancel \ 210 --s https:~/~/fts.bsc.es:8446/ \ 211 -~-~-access-token $tok \ 212 -<your_job_id>## 213 - 214 -== Notes == 215 - 216 -The configuration of the service allows to: 217 - 218 -* transfer object files with a maximum size of 5GB, note that files larger than 512MB will be stored as large objects in Openstack, namely, files are split when larger than 512 MB; each chunk is 256 MB. Note that users may still view/download the files as a whole like normal objects after the upload is complete. Information on large object support can be found here: [[https:~~/~~/docs.openstack.org/swift/latest/overview_large_objects.html>>url:https://docs.openstack.org/swift/latest/overview_large_objects.html]]. 219 -* manage up to 1000 concurrent requests. This means that up to 1000 object files, or chunks of an object file, can be sent to the storage system concurrently; 19 +
- image-20230412091641-1.png
-
- Author
-
... ... @@ -1,1 +1,0 @@ 1 -XWiki.luehrs - Size
-
... ... @@ -1,1 +1,0 @@ 1 -24.0 KB - Content
- image-20230412091702-2.png
-
- Author
-
... ... @@ -1,1 +1,0 @@ 1 -XWiki.luehrs - Size
-
... ... @@ -1,1 +1,0 @@ 1 -36.3 KB - Content