Instead of recommending that an alternative SSH service be run, provide proper instructions for running adminstrative SSH on the same port as Phabricator SSH.
Details
- Reviewers
epriestley - Group Reviewers
Blessed Reviewers
This is the same configuration that I use on my dev VM.
Diff Detail
- Repository
- rP Phabricator
- Branch
- master
- Lint
Lint Passed - Unit
Tests Passed - Build Status
Buildable 7400 Build 7840: [Placeholder Plan] Wait for 30 Seconds Build 7839: arc lint + arc unit
Event Timeline
overall seems reasonable, and the /etc/shadow issue predates your change. This sort of instruction would also make it easier to use when firewalls are in the way
src/docs/user/userguide/diffusion_hosting.diviner | ||
---|---|---|
108–112 | It would be good to mention that not all installations have /etc/shadow |
I think there are three generally reasonable ways to configure this:
- Use different ports, with VCS access on a nonstandard port.
- Use different ports, with operations access on a nonstandard port.
- Use the same port, with "Match User".
I don't think (3) is universally desirable. In particular:
- It carries the highest risk of locking yourself out of the box while trying to configure it, since you have to restart the real SSHD to test changes.
- It prevents you from firewalling the administrative and VCS ports separately. I think most serious setups should not allow public-facing operational SSH access.
- Although this isn't a concern in most cases, it prevents you from using a sshd with a ForceUser directive like we do in the cluster to get GitHub-like behavior where the user is ignored.
So (3) seems worse for less-sophisticated users (easier to make a serious mistake) and also worse for more-sophisticated users (they are more likely to want to separate operational and VCS access with network-level rules).
I'm OK with expanding the documentation to include (1) and (3) as options, but if we're recommending only one, I think (2) is still the best one to recommend.