Remove horrible old repository edit controller
c585f97e90d2
Actions

Description

Remove horrible old repository edit controller

Summary:
Ref T2231. I didn't port these options over, so they're still supported but have no edit UI:

Pull Frequency (confusing/not useful, I think?)
Default Owners Path (probably used only by Facebook and only in the E repository)
Show user in public repository URL (probably mostly obsolete with hosting?)

We can add those back if users notice, but they seem like the three least useful options so I'm going to see if we can get away with removing them.

Test Plan: Clicked "Edit" from Repositories, got kicked into the nice new Diffusion edit UI instead of the old one.

Reviewers: btrahan

Reviewed By: btrahan

CC: aran

Maniphest Tasks: T2231

Differential Revision: https://secure.phabricator.com/D7410

Details

Provenance

epriestley

Authored on Oct 25 2013, 11:22 PM

Reviewer

btrahan

Differential Revision

D7410: Remove horrible old repository edit controller

Parents

rP6bdedfe679fa: Add ObjectBox to Condiut Cert Page

Branches

Unknown

Tags

Unknown

Tasks

Restricted Maniphest Task

Event Timeline

It would be nice to have the option to reduce the pull frequency of individual git repositories. I noted that, the PhabricatorRepositoryPullLocalDaemon generates quite a few queries to MySQL along with git pull every 15 seconds. When running Phabricator on AWS EC2 EBS instance, the daemon constantly generates significant EBS IO requests.

Can you quantify "significant"?

Is the repository bare, or is it an older repository which was cloned with a working copy?

The git repository has 2 branches with less than 50 commits and is static during the test. The repository is located on the same physical disk as Phabricator with three running daemons,

1x PhabricatorRepositoryPullLocalDaemon
1x PhabricatorGarbageCollectorDaemon
1x PhabricatorTaskmasterDaemon

The average sustained OPs/s is 6 measured on the EBS volume. I suspect this would grow linearly as more git repositories are added for monitoring.

How about this?

Is the repository bare, or is it an older repository which was cloned with a working copy?

For what it's worth, this server had 14M IOPs in October, 17M in November, and 19M in December, all of which are about equivalent to 6 IOPs/s. Somewhere in there we switched from remote to hosted repositories and we've always had 6+ repositories, so that implies:

IOPs/s are not tightly related to pulls; and
IOPs/s are not linear with the number of repositories.

These IOPs cost about $2/month, too, which is <1% of total EC2 hosting costs.

The remote repository is bare but not self hosted by Phabricator.

Thank you for taking time to look into this issue of mine. Glad to know the disk IO caused by PhabricatorRepositoryPullLocalDaemon will not contribute much to the monthly AWS bill.