Upgrade Experience

From GridPP Wiki
Jump to: navigation, search

This page describes the points for the upgrade schecdule

Sites Upgrading, their previous state and reasons to upgrade

  • UKI-SOUTHGRID-RALPP
    • Running pnfs
      • Slow performance
    • dCache 1.9.4-X
      • This was not a golden release
  • UKI-LT2-IC-HEP
    • Running pnfs
      • Slow performance
    • dCache 1.9.2-X
      • This had been stable but unsupported for approx 6 months
  • Aim had been to deploy golden release before data taking
    • Data taking and release in production of a golden release both kept slipping

Ugrade results

  • Upgrade went well
    • (This was unexpected!)
  • Both sites main issue was with the postgres database upgrade for Chimera.
  • Site tested upgrading procedure before hand on spare hardware.
    • Site tried upgrade dCache central servers and to convert from pnfs to chimera at the same time with problems.
    • Site found tracking upgrade issues decoupled if only upgrading one conmponent at a time.
      • Recommend not to upgrade services and database upgrade at the same time.
    • Testing Good
    • Experience shared with other less dCache experienced Sys admin at other sites
      • (Staged rollout as a concept was a good idea)
  • Overall upgrade proceeded with fewer unexpecte breakages of services
    • Midleware itself more stable and intuitive
      • One upgrade was enacted by a new dCache sys-admin
      • Testing of upgrade by experienced sys-admin on spare hardware invaluable.

Site View on improvement in Results

  • To be completed