ClientAliveInterval not set in gitlab-shell in kubernetes
As discovered in https://gitlab.com/gitlab-com/gl-infra/infrastructure/-/issues/12016#note_599975422, we used to set ClientAliveInterval to 30 seconds for sshd-git, to send protocol level pings across otherwise idle connections. The gitlab-shell chart does not currently allow this to be configured, so it defaults to 0 (no keepalives). This means long running connections that are doing client-side things (like compressing objects before sending) can run into the 90s haproxy timeout_client setting. If we had ClientAliveInterval set to 30, there would instead be 2-3 pings over that period keeping the connection alive.
This issue is to track the co-ordination of a charts MR to allow that value to be configured, and rolling out our specific configuration once available.