9th Tier-3 Steering Board meeting
Venue
When: Friday, 7th Dec 2018, 10:30-13:30
Location: ETH Hönggerberg, HPK D32
Agenda
- User home
- Status: ~ 75 active users use ~7TB of home (and backup) on ZFS. Lack of space during Jan-June 2018
- Plans: integration with PSI nfs home service [1]
- Storage
- Status:
- dCache upgrade to 3.2
- replacement old NetApp/SGI with 4 Dlaco servers ~ 1PB
- Plans:
- add ~ 1PB with next budget
- connect NetApp to a newer fileserver
- Batch system:
- Status:
- SGE cleaning from old queues (sherpa)
- adjustment of scheduler policy for using Parallel Environment (PE) (smp)
- Plans:
- add WNs with both budget 17-18 and 19-21, starting from Jan-Feb.19
- T3 news - GPU for machine learning used by pilot users
- Plans: add 1 more node in 2019/Q1
- Perspectives/Updates:
- Central jupyterHUB
- Documentation: Storage, etc.
- Infrastructure: replacement of very old t3 switches
- Organizational points: re-arrangement of physics groups to 3 main "institution-wise" like psi, ethz and uniz
References:
[1]
While since long there was permanent lack of space on t3 home server (t3nfs01, ~7TB),
the way to resolve the problem is to migrate to a new home location on PSI standard nfs facility /t3home.
/t3home is mounted already on UIs and WNs since June 2018 in addition to current /shome and is in use by test users and newcomers.
The new default quota is 10GB to keep scripts, documentation, stuff like this.
It's configured currently with 1 daily, 1 weekly and 1 monthly snapshots (on a specially reserved PSI nfs backup volume).
In spite of the big gap in quota sizes current movement is easy: all user data from /shome is still FULLY AVAILABLE and usable.
The next possible step is to re-use current backup space (t3nfs02, ~20TB). One might append it to /shome and use combined volume (7+20 = 27TB) as a "scratch" for big size data.
--
NinaLoktionova - 2018-11-22