use -- for all the options of the exporter configuration

minor update on documentation

• D7068 deployed and applied on the workers:

root@pergamon:/etc/clustershell# clush -b -w @workers -w worker17 -w worker18 "set -e; puppet agent --test"
clush:  0/31
clush: in progress(31): worker[01-18],worker[01-13].euwest.azure
---------------
worker01.euwest.azure
---------------
Info: Using configured environment 'production'
Info: Retrieving pluginfacts
Info: Retrieving plugin
Info: Retrieving locales
Info: Loading facts
Info: Caching catalog for worker01.euwest.azure.internal.softwareheritage.org
Info: Applying configuration version '1643885189'
Notice: Applied catalog in 11.65 seconds
...
------------
worker18
---------------
Info: Using configured environment 'production'
Info: Retrieving pluginfacts
Info: Retrieving plugin
Info: Retrieving locales
Info: Loading facts
Info: Caching catalog for worker18.softwareheritage.org
Info: Applying configuration version '1643885204'
Notice: /Stage[main]/Profile::Mountpoints/Mount[/srv/storage/space]/options: options changed 'rw,soft,intr,rsize=8192,wsize=8192,noauto,x-systemd.automount,x-systemd.device-timeout=10' to 'ro,soft,intr,rsize=8192,wsize=8192,noauto,x-systemd.automount,x-systemd.device-timeout=10'
Info: Computing checksum on file /etc/fstab
Info: /Stage[main]/Profile::Mountpoints/Mount[/srv/storage/space]: Scheduling refresh of Mount[/srv/storage/space]
Info: Mount[/srv/storage/space](provider=parsed): Remounting
Notice: /Stage[main]/Profile::Mountpoints/Mount[/srv/storage/space]: Triggered 'refresh' from 1 event
Info: /Stage[main]/Profile::Mountpoints/Mount[/srv/storage/space]: Scheduling refresh of Mount[/srv/storage/space]
Notice: Applied catalog in 19.67 seconds
clush: worker[01-18] (18): exited with exit code 2

completely remove the mountpoint to remove as the mount class
is not doing the cleanup when it's declared as absent.

There are also:

• a LB for the postgresql replicas : https://portal.azure.com/#blade/Microsoft_Azure_Network/LoadBalancingHubMenuBlade/loadBalancers (swh-postgres-public)
• 2 cosmos db for provenance (almost empty)

a few minor remarks

Thanks for making mypy happy ;)

LGTM, it's not really ideal but it avoid to break the current subnet spliting we are using

Unfornunately I didn't find any tool supporting natively the backup of postgresql through zfs snapshots.
We have a couple of alternatives:

• Use borg to to implement a clean backup of the postgresql data directory based on hooks [1]
• Use a tool like pgbackrest[2], the backup type and storage target need to be choosen (S3, storage server, ...)
• Implement a backup based on zfs snapshots, which should not be too complicated but we need to manage all the plumbing to deal with full/incremental backups and the history cleaning

related to T3889

• rebase
• update according the feedback

LGTM

thanks, just a small inlined remark regarding the httpie package installation

rebase

rebase

fix the typo on the getty hostname

• Rebase
• Update the kafka credentials part to use the new user management script
• Add a section to describe the username pattern
• Add a section to describe how to create the read-only storage credentials

rebase

add the puppet header on the script

• Install the scripts for all the environments in getty, the journal orchestrator;
• as the cluster configurations are now global, it impacts the consumer group exporter. It make sense to move it from storage1 to getty to also centralize this part (FW rules will need to be adapted accordingly);

Yes it could, It will be cleaner, I will update in this way.
I took a shortcut estimating we should know which broker is up or not, to avoid to have to build the connection string :)

rebase

The ack late propery is deployed in staging and production.
The number of scn in the 'scheduled' status should decrease.

rebase

fix according the review

• rebase