[tahoe-dev] 回复: Re: [tahoe-lafs] #643: Automatically schedule repair process (and backups?)

[Zooko O'Whielacronx]

On Thu, Aug 15, 2013 at 5:41 AM, han zheng <z.han0903 at gmail.com> wrote:
>
> I have done something about this. I have written some code for tahoe to automatically schedule repair process.

Hooray!

> the basic idea like this: run a monitor service when tahoe client runs, detect file's health'state when downloading a file, if the file is unhealthy(the valid shares < a threshold number), then record this file, and at a scheduled time(maybe at midnight), run a repair service to make this file healthy...

This sounds like a good approach!

As a user, I would want a way to see information about this: how many
files have been detected as unhealthy, how many got repaired each
night, what times the repair process is running, etc. That sort of
transparency needs to be included in the UI.

> My problem is how to make a test to this? I think this way can save much resources consumed by a repair process. I know there are simulators in the source code folder of tahoe, but I donnot know how to use it to test this repair process. I want to know whether this way to repair files can provide a good way to save repair overhead compared with original way and keep a high availability at the same time.

Have you seen "How To Write Tests" on the wiki?

https://tahoe-lafs.org/trac/tahoe-lafs/wiki/HowToWriteTests

Please follow those instructions to create a new test named
"allmydata.test.test_autorepair".

Also, please create an account on github.com, and fork this repo:

https://github.com/tahoe-lafs/tahoe-lafs

And push your patches to your fork, and put the hyperlink to your fork
into a comment on https://tahoe-lafs.org/trac/tahoe-lafs/ticket/643 .

Thank you!

Regards,

Zooko