[prev in list] [next in list] [prev in thread] [next in thread]
List: dragonfly-commits
Subject: Re: cvs commit: src/sbin/hammer cmd_mirror.c cmd_pseudofs.c hammer.8
From: Michael Neumann <mneumann () ntecs ! de>
Date: 2008-07-31 7:25:03
Message-ID: 489168CF.9090600 () ntecs ! de
[Download RAW message or body]
Matthew Dillon wrote:
> : Log:
> : HAMMER Utilities: Streaming mirroring!
> :
> : * Add a new streaming mirroring feature called 'mirror-stream' which works
>
> Please note that this is not quite ready for testing yet. There is
> a race between normal filesystem writes and the mirror's B-Tree scan
> which can cause a DATA CRC and other assertions.
>
> In fact, the only known bugs in the filesystem at the moment, assuming
> I fixed that blocking bug reported on the lists, is due to buffer
> cache aliasing between vnode-mapped buffers and device-mapped buffers.
> I think the cluster_read I do on the device-mapped buffers is causing
> the headache.
>
> I hope to have the bug fixed by tomorrow evening.
>
> The new bandwidth-limited streaming mirroring feature is going to be
> very spiffy! It's working except for the above bug!
>
> It will make it possible to have a streaming mirror continuously
> running in the background but not eating much in the way of system
> resources. So instead of batching your backups you would be able to
> maintain a continuously running backup. It should also work well for
> sporatically connected boxes, like a laptop, where you want it to
> maintain a backup automatically whenever it happens to have internet
> connectivity, but also not eat up too much of your WIFI bandwidth.
That's pretty cool! One question though: If I'd like to have the exact
same snapshots on each machine, is that as easy as just "copying" the
softlinks? That is, are the transaction ids (those @@0x... numbers :)
identical on the master and it's mirrors?
Regards,
Michael
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic