Re: Reliable WAL file shipping over unreliable network

Really? I think you really need to think about it. You are dealing with a race condition.

While your though process might seem to work and it will appear to work; however in the end it will fail — like I said; I’ve seen it fail first hand.

You are dealing with a race condition involving filesystem metadata. If you choose to not use “-c” option then; all I can tell you is that you have been warned.

> On Feb 28, 2018, at 1:26 PM, Dianne Skoll <dfs(at)roaringpenguin(dot)com> wrote:
>
> On Wed, 28 Feb 2018 13:15:48 -0500
> Rui DeSousa <rui(dot)desousa(at)icloud(dot)com> wrote:
>
>> Sure it does… what happens if the archive server gets full? There is
>> a situation that can fool rsync into thinking it was successfully
>> transferred as rsync only relies on filesystem metadata to verify the
>> file is different.
>
> That's incorrect. The original transfer will fail and rsync will remove
> the temporary file. It doesn't rename the file to the final destination
> until and unless the transfer is successful.
>
> "-c" is only used for files that might change multiple times per second;
> since most file systems have only 1-second granualarity in file timestamps,
> you can't rely on a file being identical if the size and timestamps are the
> same.
>
>> The “-c” option will validate the file using a checksum and not just
>> rely on just filesystem metadata.
>
> That's incorrect. "-c" applies only when rsync looks at existing files
> to see if they differ. It does not affect what happens when rsync actually
> copies the file data over; that's always verified with a checksum.
>
> Regards,
>
> Dianne.
>