[tools] Retrying failed rsyncs w/mrepo?

[Dag Wieers]

On Tue, 11 Dec 2007, Wil Cooley wrote:

> I frequently have failures rsyncing because of temporary problems with
> the mirror I am using, generally due to load. Other than changing
> mirrors (I've changed several already; it happens often enough that it
> seems like it would be a game of whack-a-mole), would you consider
> implementing a retry mechanism? If there is interest, I might take a
> shot at implementing it; I have already done something similar with a
> Perl module I wrote. I have precious few opportunities to practice my
> Python skills.

The problem with a retru mechanism is that you need to know what happens
before retrying. rsync can fail for many reasons, and some of these
reasons are not fixed by retrying (like permission errors).

Retrying in such cases makes little sense. The problem is thus, how can we
now for sure that rsync had a problem that is worth retrying for.

I would prefer to look for a solution within rsync itself, and if that not
exist look with the rsync developers if some of the problems can be
avoided (by eg. changing timeout parameters and such).

We had a discussion wrt. this with lftp as well and I am not convinced
that mrepo should do process management or fault-recovery, mainly because
mrepo has little knowledge of faults in the first place and you cannot
recover something you have little knowledge about. (It might be for the
wrong reasons or with no impact)

In the lftp case, it was simply a bug in lftp.

-- 
--   dag wieers,  dag at wieers.com,  http://dag.wieers.com/   --
[Any errors in spelling, tact or fact are transmission errors]