NDGF All Hands 2

Europe/Copenhagen
room KBJ302 (HPC2N)

room KBJ302

HPC2N

KBC building, Umeå University, Umeå
Description

NT1 staff, and local NDGF site sysadmin meeting.

The session before lunch are for internal topics to the NT1. The local site admins are welcome to attend, but please keep snoring to a minimum.

The session after lunch are for the local site admins. This is your chance to learn from your colleagues from around the Nordics. Topics for discussion do not need to be strictly tied to T1 operations, but can cover any topic of interest to running an HPC shop.

Registration will close Friday at noon. You are still welcome to show up, but there will not be Kanelbullar for you.

The day after is a combined Nordugrid developer and NDGF sys admin workshop. See here for details: http://indico.lucas.lu.se/event/1020

Getting to the venue is a 25 minute walk from downtown where the hotels (and dinner) is, or a matter of taking bus 2,5,8 to Växthuset. Buses accept credit cards but not cash, enter by the front door and pay the bus driver. Main entrance with signage to the KBC building is on the west side.

Travel: Airport is a short bus or taxi ride away from both central Umeå with hotels and the University. It is also possible to do as a long walk (4-5 km). Taxis offer fixed prices which should be around 200-250 SEK, payable by credit card. At the airport unbooked taxis occupy the lane furthest away from the exit.

Trains stop at either Umeå Central at the city center or by Umeå Östra next to the hospital and university campus, same applies to long distance buses.

To the airport: Leaving (getting on the bus or in the taxi) one hour before departure should be sufficient, 45 minutes if you are not checking any bags.

https://wiki.neic.no/int/NDGF-AllHands2018_2_minutes

There are minutes attached to this event. Show them.
    • 09:00 12:00
      NT1 staff internal meeting
      • 09:00
        dCache preprod monitoring 20m

        Do we need more probes for the pre-production dCache setup on mordor?

      • 09:20
        New Ubuntu release readiness and upgrade plan 30m
      • 09:50
        ood/ooc scheduling 15m

        Please prep the common calendar the week before. Chrulle will then come up with a suggested schedule ahead of time. If you are not OoD or OoC you have an early break :)

      • 10:05
        Publishing our scripts 20m

        Discussion on publishing our ansible scripts etc on github.

      • 10:25
        Coffee 20m
      • 10:45
        Kafka for everything? 30m
        1. What can we put into kafka in addition to dcache billing.
        2. What is our strategy towards logstash if we put everything into kafka:
          a. Ditch it and install es-consumer that feeds data directly to the ES.
          b. A separate instance on dedicated machine that queries kafka on different topics and submits them to the ES.
          c. Other.
      • 11:15
        Clom & Zanak replacement 20m

        Service for C&Z will run out on 30 September 2019. It is time to start deciding on and procuring the replacement. Discussion on sizing our production Ganeti and Postgresql cluster.

      • 11:35
        Postgresql topics 20m

        OS / FS tuning: overcommit, dirty pages etc.

        How are we handling DB backup today. Should we look into alternatives?

      • 11:55
        Lunch venue decision 5m
    • 12:00 13:00
      Lunch 1h
    • 13:00 17:10
      NDGF local sysadmins meeting
      • 13:00
        Site news and plans 1h

        Sites will prepare a short presentation on news and plans.

      • 14:00
        LHC news and future 30m

        Oxana will give a presentation on LHC news and lead a discussion on future requirements.

        Speaker: Oxana Smirnova (NeIC / Lund University)
      • 15:00
        Coffee 30m
      • 15:30
        Future tape performance 20m

        Future requirements for tape performance, with implications and discussions on tape pool hardware

      • 15:50
        Ceph for compute and storage 30m

        Discussion of experiences with using Ceph and ARC together and for using ceph with dcache.

      • 16:20
        Accounting/reporting of local jobs 20m

        Discussion on including local jobs in reporting.

        Speaker: Mr Ville Salmela (csc.fi)
      • 16:40
        Disk pool requirements 20m

        The usage of ATLAS and ALICE pools today look pretty different. Is the usage going to diverge further? Or is it going to converge? Are we going to need smaller pools more suited for random IO?

      • 17:00
        Next NDGF All Hands 10m
    • 18:30 20:30
      Dinner 2h