OpenStreetMap Foundation, Operations Meeting* - Agenda & Minutes
Friday May 22nd 2020, 18:00 London time
Location: Video room at https://osmvideo.cloud68.co
- Please note that this was not strictly an OWG meeting.
- Tom Hughes (OWG)
- Paul Norman (OWG, board)
- Emilie Laffray (OWG)
- Michal Migurski
- Guillaume Rischard (board - joined for a few minutes)
Minutes by Dorothea.
Topics were reordered during meeting.
<link to be added>
Open Ops Tickets
Review open, what needs policy and what needs someone to help with.
OSQA currently used
- inactive upstream for long time.
- Python/Python2, which goes away.
1 - ASKBOT (OSQA fork)
- Github not the most active. Not big number of contributors (1 in last year but more than question2answer).
- Migration probably doable but might be an undertaking.
- We did try to reach out to askbot dev (might have done through 3rd party).
- Migration path probably more straight than question2answer.
- Developer offers consulting.
2 - Question2answer
- Repo not very active either - last commit Feb 2019
- Bugfix branch https://github.com/q2a/question2answer/issues
- We'd lose account integration (3rd party control but something less to maintain).
- We'd lose historical content.
- Existing curation reasonable.
- URLs relatively sane for archiving purposes.
- Separate issues of archiving old content and switching to new content.
- Put current content on archive, preserve links (Fedora did something similar on ask.fedoraproject.org: was askbot, now Discourse on same url).
- Archive in static form to be future-proof.
- Will require some trust / anonymisation of database.
- No passwords contained, as authentication via osm.org, but has email addresses. Will have to keep user iDs, scrap everything else (tokens).
Other points during discussion
- Django community could help.
- Michal suggested creating sandbox / site clone for testing purposes
- Need root access to play with software.
- Chef configuration existing.
- Paul to check Askbot on Python 3.0 - can contact consulting and say we're considering to use it and would like to use Python 3.0.
- Michal to look into archiving.
Not minuted due to security reasons.
Squid / nginx / GeoDNS? Is the software right? Issues and improvement.
(carried over from April meeting)
- Difficult to understand health of network.
- Squid cache miss time indicates not healthy.
- Median response time of multiple servers: up to 10s of seconds.
Under-powered machines - hard to get rid of them
- Make sure we have the capacity elsewhere: ~ five machines serving a lot of traffic in Germany. If we turn off 1-2, the rest might not have the capacity to handle the traffic.
- Donated machines.
- We can stop using them - just say they "we've tried the donated machine you've offered but it is not working with our load".
- If we don't arrange something with sponsored commercial CDN, we should look at getting some more powerful machines. We know hosting companies in FR and DE.
- We should not be pinning everything on Fastly.
Any other business
4 up and running (2 in Oregon)
Machines running old software
- Archiving necessary
- Andy is in the process of going through website issues and cloding/redirecting them to Github.
- Content has less value than help.osm.org
- No one is filing new issues that they have any reasonable chance of being resolved on trac.
- OSGEO was a big user and is moving away from it.
- Not going anywhere as a technology.
- Set date after which no new accounts will be created associated with it (no further support).
- JOSM plugin developers are creating SVN accounts and using it as a CDN.
- Commit not only the source code but also the built. JOSM then uses the SVN url to download the .jar for the plug-in during installation.
- Migrating the JOSM community is a prerequisite for shutting SVN.
- Page on JOSM wiki which lists the URL for each plug-in.
- A bare-bones read-only SVN server is not so much of an administrative burden as one being used.
- Firewall issue at UCL -where Github actions were hosted- with inbound connections at ports 1025/1026 were being blocked.
- UCL put a rule for us and was fixed.
- Blog testing is failing randomly - tries to fetch all the blogs so if any has a problem, it fails.
- Red and green alerts give inaccurate sense of how much is working versus not working.
Action item: Tom to make Chef ignore failures on blog - have to distinguish between runtime errors from more serious problems.