01.09.2023 TT Agenda - QualitativeDataRepository/TechnicalTeam GitHub Wiki

Admin:

  • Postmortem for 12.26 outage
  • Access and directions to SK and NW to reboot the server if down (seba?)

Seba updates:

Jim updates:

For discussion:

  • Matomo tracking code on prod
    • Reproduce the google analytics configuration where we are tracking clicks but not page loads, etc

Update:

  • Dev instance of Matomo setup -

  • What is the goal? - Karcher is currently importing the google analytics history to Matomo (live sync). Ultimate goal is replication of GA tracking we have developed...

    • On dev property - event triggers can be set up so we are capturing the same info on Matomo...

  • Do we want more features - such as the opt-in vs opt-out that Matomo offers in a dialogue box? A: Since we aren't implementing cookies we don't need to make users opt-in to anything.

  • https://github.com/IQSS/dataverse/issues/2917 ... Has been stuck at DV for some time. QDR would benefit from solving this issue - How big of a time commitment, and do we have any read if dataverse would accept a PR on this?

    • On QDR side - copying from OpenAire side is easy... A few hours worth of work. What is stuck is agreeing this is time to switch - We could do this and push it back

  • https://github.com/orgs/QualitativeDataRepository/projects/1?pane=item&itemId=9245809 ... Lets discuss what redesigning account management (SSO) might include, weigh some tradeoffs of different designs, and estimate time commitment.

OIDC with keycloak is potential way to maintain Shib and SSO connections - if we are logging in users through Drupal then we might have an easier handoff (essentially we move from two parallel systems to one) ... The things that might be messy are going to be allowing users to continue to create accounts managed by QDR... For HEAL - this might provide an easier handoff of users authenticating with OIDC - the session could essentially handoff a user with the trusted session id... HEAL/OIDC - Dataverse doc avail ~this week. Can start investigating Keycloak OIDC provider in front of/replacing Shib IDP for internal users. SSO depends on TBD details but could be simplified from current with new login/logout API calls in Dataverse.

Time estimate: If we're really just handing off tokens it could be relatively easy ... SPA work is contracted at Harvard, and they have a few weeks out to that being delivered... It could then take 1-1.5 months for implementing at QDR.

Notes from Seba:

  • Failure modes
    • Drupal down; AnnoRep down; DV down; some combination of those three
  • A few solutions - it could just be restarting the pipeline - SSH commands

Seba

  • No updates - shenanigans with Nagios

To work on:

  • Key rotation
  • Increasing costs for RSC - can we look at daily breakdown to make sure that the costs aren't related to

Question

  • Stage disk space ... infrequent, but a pain. EBS - can add more, or just clear out logs to give more space.
  • WARC files can all be cleaned out (Jim can delete more and we can reassess)
    • In future, we should automate the purging of archives just to make sure we are getting these cleared out...