Ubuntu20 redash

[ghost]

This PR updates salt to deploy redash on the new ocp14 server running ubuntu 20.04 #145

I have also taken this opportunity to move postgres out of docker and upgrade it (closes #299). Because of this change and specifically vm.overcommit_memory, the memory requirements on this server has increased to needing at least 6GB SWAP / 4GB RAM (from 2GB SWAP / 4GB RAM)

Few tasks are remaining before this is merged in:

• Update Documentation for onboarding.sls change
• Import the redash database and test the site
• Configure networking for the new IPv6 linode block Registry network (netplan) configuration #278
• Update the redash set up docs https://ocdsdeploy.readthedocs.io/en/latest/deploy/redash.html

@jpmckinney / @yolile , I can see redash is configured to send email through SES, what is this used for? If I import the database from ocp08 for testing will it trigger any emails?

[jpmckinney]

For the docker-compose file, the differences are:

• remove postgres
• upgrade redis and add healthcheck
• change docker-compose file version
• change docker image tag

[ghost]

@jpmckinney , this PR is ready to review.

I have...

• Updated redash to use docker_apps
• Added netplan configuration enabling us to configure IPv6 on Linode servers
• With the netplan changes, I have added configs to registry addressing issue Registry network (netplan) configuration #278
• Updated documentation around onboarding and created docs for netplan
• Updated redash version for the new stable release
• Updated redash documentation for new setup and processes.

I have also imported the database from the old server onto ocp14 and tested, confirming the app is working fine :)

[ghost]

Noting migration steps:

1. Merge PR and changes in private repo
   • https://github.com/open-contracting/deploy-pillar-private/pull/24
2. Follow Documented "Initialize Redash" process
3. Copy over database from OCP with one additional step:

sed -i 's/OWNER TO postgres/OWNER TO redash/g; /COMMENT ON DATABASE postgres/d' redash.sql

4. Update Redash DB secret key

docker-compose run --rm server manage database reencrypt "old_secret_key" "new_secret_key"

5. Update DNS

Edit: Updated steps for latest process.

[jpmckinney]

Almost there!

[jpmckinney]

I assume "latest" means "last"? I changed it to avoid ambiguity.

That said, https://www.linode.com/docs/guides/linux-static-ip-configuration/#ubuntu-1804-and-2004 puts the SLAAC address last. It also includes a private IP. I don't know if "IPv6 – Link Local" in Linode matches that private IP or is something else.

[ghost]

I assume "latest" means "last"? I changed it
✔✔

In my testing I've had lots of issues due to the ordering the IPv6 IPs. The config in this PR is the most reliable set up.
As I remember, putting SLAAC last will prioritise the other IPv6 address (from the Linode pool) over our IPv6 pool.
The server picks up the Link Local address automatically so no need for manual configuration.

[jpmckinney]

I'm having trouble finding documentation about the priority order. This thread suggests that the first entry is prioritized, not the last: https://bugs.launchpad.net/netplan/+bug/1766656

The manpage for systemd-network doesn't seem to describe any ordering logic: https://manpages.ubuntu.com/manpages/bionic/man5/systemd.network.5.html

I read the Address= description, and searched for first, last, order, priorit*, primary, secondary, additional, prefer.

[jpmckinney]

Anyhow, the logic seems to be undocumented/unspecified, so if this config works, let's go ahead with the merge. I've added comments to the file, in case we later discover documented logic.

[ghost]

Agreed - the documentation around Netplan is poor.
I'm happy that this configuration is working fine on OCP14.
I will merge this PR tomorrow morning (20 Apr 2022). :)