03.08.2021 TT Agenda - QualitativeDataRepository/TechnicalTeam GitHub Wiki
People
@nniicc @adam3smith @stahs @qqmyers
Discussion Topics
Admin
- TAB meeting was good - longer discussion of DR recovery and backup
- Its Reserved Instance time of the year - SK check with SO on current needs
Seba & Jim Coordination
- Plan for automated cleaning of older files in /tmp (which tool, who does the work)
- Should stage be updated to match dev (with ~5.4 changes) or just style fixes that could go to prod ~now?
- discuss DataCite 4.x support (question from Sebastian)
Seba
- Item
Jim
- Report '
- AnnoRep Update from SK (keeping Jim in loop)
Assigned Tasks / Decisions
- SK - Reserved instances renewal - Same as 2020
- SO - Look into version control on S3 buckets
- SO - Look into /tmp tools for purging so we don't run out of disc
- SK + NW - find out date for downtime so Seba can prep EBS
- JM - Put ~5.4 on stage along with style change issues
Notes
Seba work accomplished:
-
Seba adjusted scripts to send backups to ICT + Verified (this is including DV databases from S3). Uses similar logic of duply we had originally, just updated.
-
ICT backups are verified and working as of Sunday 03.07
-
This will be documented in backup page on Infrastructre docs in Github...
-
The backup ICT vpn set up is non-trivial for Seba locally
-
And the backups are not as accessible for entire team - but we CAN get them ... Will work going forward on making these more accessible and documenting this ...
Renew reserved instances ...
- Should we just renew blankly or is there anything to adjust
- If you buy RI of an instance class - the size doesn't matter that much (C5, XLarge or 2Xlarge...)
- Current -T1 and T3 are burstable and cheap ...
-
- Metrics we are nowhere near performance bottlenecks ... There is nothing cheaper either so we're good.
Temp repo cleaning
- What should we use in order to do these purges
- And who should hold the bag...
-
- Production usage went up to 74% ... So not immediate because we can manually delete...
-
- When does storage alert kick in? 80%
EBS situation...
- Combining EBS volumes are the next step
- Let's discuss when can we take downtime - Probably around 30 min ...
- Cost of backups ... Writing out full s3 copies or are we using duply? There is no change to Duply config for ICT backups...
-
- For S3 Dataverse backup - there is no increase or movement in cost (that we observed)
-
- DR backups ... It's a pool from S3 that we are going to use...
Is there any S3 built in backup that we can use to do backup in addition to ICT ...
- Having a warmish back up on ICT would add layer of security
- Rsync style backup would be helpful there...
Jim - seven 9's worth of backup - and there is versioning that you can put that into place - all of our files are write once
SK - is there an ability to perma-delete (for use case when we download, delete, and reupload)
Are we paying for additional S3 storage ... ?
- We don't think so - but we should check on the version control for the bucket