Presentation is loading. Please wait.

Presentation is loading. Please wait.

Farida Fassi, Damien Mercie

Similar presentations


Presentation on theme: "Farida Fassi, Damien Mercie"— Presentation transcript:

1 Farida Fassi, Damien Mercie
vendredi 20 juillet 2018vendredi 20 juillet 2018 T1 activity report Farida Fassi, Damien Mercie

2 Current disk allocation Disk deployment dCache setup
Outlines Current disk allocation Disk deployment dCache setup CMS computing visit and actions triggered Brief review on the CMS activities

3 dCache: current disk allocation
Import pools: 161TB, Where 120TB allocated for Space-token CMSCUSTODIAL. This pools is the buffer for all the incoming data: from T0, T1s, T2s and T3s. Export pools: ~27TB, buffer that holds the outcome data, for both T1 and T2. - Reprocessing pools: ~327TB New Configuration: data flagged at import pools will be kept on buffer, more details later on Unmerge pools (T0D1): 228TB for T2 and 30TB for unmerge activity Due to technical limitation the T2 disk pools and unmerge pools have to be the same pools. CMS support has to be aware of the status of this pools permanently since they hold the output of a chaotic activity, plus reprocessing a big pools can reduce the impact in case of space issue

4 Disk allocation for T2 and T1
T2: prod: to be increased by 100TB (To be checked) T1: prod pools, unmerged area was increased by 32 TB import pools increased by 62 TB Debug pools increased by ~2TB 200 TB need to be installed in the Reprocessing pools

5 dCache pools setup HPSS Analysis Jobs PhEDEx transfer T0, T1, T2 rfcp
Unmerge Pool, T2 POOLS Import pools T2 pools Unmerge Export pools Re-processing pools Prod-Merging Jobs Reprocessing Jobs Analysis Jobs Production Jobs January 09

6 Preprocessing and Presaging
CMS requests that real and Gen datasets has to be kept online during one year, to guaranty a fast access to data for reprocessing purpose. For that purpose an implementation was done to tag this kind of data once they lend on the import pools. The data flagged as such will be moved to the Reprocessing pools to be kept there, while a copy of the same data will be migrated to HPSS This option of keeping data on Reprocessing pools was chosen by the fact that is more safety and transparent than the one of keeping data on the import pools. Presaging: A prestaging test using “central srm-bring-online” script was done with Treqs. The test was successful, and the average transfer rate was 106 MB/s. SRM script will be used for automatic prestaging for CCIN2P3 An automatic mechamis was put in place for cleanup unmerge area

7 CMS computing visit: Actions on CCIN2P3 (1)
Since the CMS visit to CCIN2P3 on 23th October several actions were triggered. Three meetings were organized (attended by Fabio and Farida) to establish the list and follow up all the progress . The goal is to bring CCIN2P3 to a stable and a competitive T1 site. 17 actions were considered and split in three categories depending on the issue urgency: high, medium and low. The actions cover mainly the following: Transfer in both instances Prod and Debug Specific services deployed at CCIN2P3 such as Phedex and Squid Storage configuration, in particular dCache Reprocessing and Prestage More details are on:

8 CMS computing visit: Actions on CCIN2P3 (2)
The monitoring of the Voboxes was improved. This includes the CMS-specific services (all phedex agents, Squids) and the machine level parameters More details: Nadia’s talk Transfer in Debug instance was improved considerably by the following intervention: Re-configuration of the dcache pools done, this includes increasing the disk pools used for this activity up to 5 pools of 5TB total size. Reconfiguration of the TFC to the new dCache space-name The configuration of each pool was reviewed to support 30 concurrent transfer The FTS configuration was reviewed to handle more than 60 links for this instance The effort has been done in the IN2P3<->start channel since the diagnostic and issue detection is not easy in this case. An automatic clean up for the Debug import area was setupped Transfer in PROD instance, the agents configuration was reviewed and updated deployment and configuration of the prestage agent for Transfer

9 Brief activity report

10 Transfer Prod: * -> IN2P3
F. Fassi 02/12/10 10

11 Transfer rates Prod: * -> IN2P3
Damien proxy FNAL import: main issue Timeout : it was increase up to 10000s  stable situation. Files size up 40GB - Other issues that impacted the import: Schedule Downtime on 4th and 25-26th January SRM issue on 12, 11 and 17th January Damien proxy issues F. Fassi 02/12/10 11

12 Transfer rates: IN2P3->* (PROD)
SRM issue on and 17 Few activity + SD on 4th Issues at the site that import 25-26th SD F. Fassi 02/12/10 12

13 Transfer rates: * -> IN2P3 (Debug)
Proxy issue F. Fassi 02/12/10 13

14 Transfer rates: IN2P3->* (DEBUG)
More links were added F. Fassi 02/12/10 14

15 Reprocessing Few activity since the end of year 2009

16 Comments/questions are welcome


Download ppt "Farida Fassi, Damien Mercie"

Similar presentations


Ads by Google