RAT Unit Meeting -- 27-01-2020

Projects

CompProj:379 Review of DICE desktop platform

• gdutton: completion report in progress
• gdutton: create new project once report done: purpose has changed to wider than original remit

CompProj:392 Live Chat Service

• discussed with CSO's, many opinions, another non-queued interruption would be a problem
• looked at some packages
• can we use teams - this has mobile app, web client is fine on linux, but automation may be needed to make this work in ways like live chat, including returning to RT
• gdutton: interim report on status and options pending
• gdutton: revisited some work on Mattermost for other purposes
• low priority

CompProj:417 Roles Management

• gdutton: met Toby for remaining deliverables, report written. Notes at gdutton:EntitlementTools
• finish project when roles supremo page is sufficient

CompProj:420 Hadoop

• timc: check if dpia needed, yes
• we need to decide if we can auto-delete and/or retention policy
  • could have two stage process i.e. archive first then delete, with email to say accessible but not run jobs, will be deleted after set period of time otherwise

CompProj:463 Merged MLP/MSC Teaching Clusters

• script done to generate auto.homes file driven by caps - important so we can separate users onto the correct filesystem, teaching or research
  • add ability to get config from file
  • now shifting people about, along with the amount of data involved means this will take a while to do
  • done using rsync
• script managing home directories nor running in reporting mode for safety
• working on component, hopefully LCFG mangaeable
• working on DPIA

CompProj:464 User access to last/ps/w etc

• gdutton: concrete proposals report continuingTHIS WEEK

CompProj:470 Personalised Portal Page for Academic Staff

• ongoing, "select" academics testing; need to demo for admin; core development is done; many reports to add or transfer from Portal but that is not part of this project
• continue on documentation
• status: very soft launch!

CompProj:506 Teaching Software 2019/20

• Node.js added (RT:99400)
• S2: nbgrader remains

HTTP -> HTTPS check

• rwb to investigate, piggy-backing on neilb efforts
• Nothing major anticipated.

CompProj:539 User Facing MHR Data Asset Register

• gdutton: met with cms to capture requirements and consider development for users
• will stall for the time being

GPU Approved Supplier Procurement

• ITT went out on 24th Jan, deadline is 14th Feb, evaluation 15th to 21st
  • make sure procurement give us all tender related documentation from suppliers
• some items will have to be purchased as "call-off"
• delivery scheduled end March.

Misc Development

• continuing on migrating some TSP processes
  • still todo form 2 handling, some tweaks requested following review meeting
  • gdutton: other TSP enhancement
  • timc: go over changes with gdutton

• RT lifespan/retention plus other things - discussed at ops
  • DPIA should precede these changes.
  • add accounts by ldap
  • look at merging emails - needs QA process
  • look at purging and retention rules
  • run pre move to postgresql scripts

Operational

• 7.6 upgrades
  • CDT VM host is done

• h/w security decommission
  • wasserboxer - move bridgeport off - flapjack due to be re-installed, due this week
  • fondant - off, needs to be physically removed
  • arcsim - ssh access only (iptables), also wants a VM server for arcsimvm1 (commonrail)
  • bocian/blanik/karenin
    • timc: all ssh access only while migration completed by Rob, manual iptables configuration
  • henwen + wilbur - due for removal now, rack space is precious

• GPU networking
  • iainr: discuss infrastructure with gdmr/idurkacz
    • add to spending plan if appropriate
    • 10gb cards now pulled
    • 10gb switch now populated with 10gb modules

• iainr has got power figures from "hannah", max power draw about 8.5A with all GPUs running but if running CPUs power draw goes down
  • data can be collected live from Dells so can get historical maximum, need to check if higher than aircon maximum

• doing DPIAs:
  • we need to do all our services, in approx priority order:
    • Webmark
    • Theon
    • TheonPortal
    • RT4 (can use some of Unidesk replacement one)
    • ProjSubs, Projects-Archive
    • DPMT
    • Slurm
    • Hadoop
    • License server logs
    • Lab exam?
  • aburford: need to do one for ProctorU
  • timc: doing WhosOff (probably not)
  • privacy statement for:
    • Codegrade - draft with Rena, legal dept have written DPIA
    • Webmark

• tophat attendance * privacy question, may not be specifically tophat, sitting with IS for comment

• looking at live capture (transcript) options
  • live broadcast audio much better, latency in transcription may be significant
  • doing more testing with disability office
    • issues with videoconferencing hampering efforts; loan hardware should help
  • likely to be a compromise approach
  • Nick helping with testing, gone back to disability office to suggest purchasing iPad and microphone
    • we have an iPad we could use for testing - ask Alastair

• lennoxtown still with Novatech, some response since 1st Jan
  • re-prod

• ubatuba also still broken (2 GPU failures)
  • will wait on lennoxtown and then do a GPU swap to test

• laphroaig still falling over weekly(?) in use
  • also partially knocks over BMC, further investigation required
  • fell over again - using an ordinary drive for OS and see what happens
  • check for newer firmware?

• PG v12 - no OIDs anymore, affects TheonUI
  • synthetic oid column is a minimal safe fix

• Theon (meringue) downtime
  • timc: need to deal with Learn embeds - specific 403 for learn subdirectory would be useful
  • would be nice to monitor postgresql configuration failures at component level -- gdutton to bugzilla

• postgresql access
  • local trust access allows any authorised user to login as any user account
  • this does not affect our general usage but in the context of a further authorisation failure we would have been at risk
  • remove localhost IP access (we don't need to use it)
  • move all other trust types to peer, adds a further level of security
  • change permissions on socket to limit access by unix uid
  • should probably move nagios to gssapi - need a little bit of work

• shuffling data from teaching cluster to research cluster
  • iain will do a few and write procedure so richard can then assist

• pgteach move to teacake / v11
  • done

• mattermost trial
  • vs. Slack, vs. MS Teams
  • going to use a team created on the EPCC service for trial
  • Inf VM seems to be next best option if we must host it.
  • SSO/Authentication can be added to Open Source edition.

• removed final FLEXlm from two machines, including legacy matlab and Simics

• report Rob's VM issues following edge switch reboots
  • bonding protection caused hosts to not see network failure
  • George has proposed solutions written up in infrastructure report for next ops meeting
  • gdutton: bouncing guest media state should avoid reboot, liasing with Rob to test

• ILCC cluster
  • all these machines should be under Slurm by Wednesday
  • likely to be purchasing more

AOCB

• Nagios RAT Issues
• BAD-RAT tickets