On 10 May 2016 at 07:15, Eric Drechsel <[email protected]> wrote:
>
>
> On Sat, May 7, 2016 at 7:05 PM, Alok Parlikar <[email protected]> wrote:
>>
>> Probably an FAQ, and I did browse the group archives, but might still need
>> a little handholding in setting up my workflow.
>>
>>
>> I have two linux laptops, an android phone, a RaspberryPi and a host with
>> linode. The linode box currently has owncloud installed. My backup workflow
>> currently just has duplicity with external hdds.
>>
>> I'd like to switch to camlistore -- looking at all the docs and videos, it
>> was easy to make up my mind. But I'm stuck as to how to actually set up my
>> workflow.
>>
>> (1) Which machine(s) should run camlistored? One of my laptops? The
>> linode? Or should I switch to GCE/AWS instead of linode and use cloud
>> storage for backup?
>
>
> I'd recommend running a primary instance on an always-connected, always-on
> host.
>
> I have in the past run a secondary instance on my laptop for local access,
> but there is not yet support for syncing subsets of your data (for example
> your "Documents" folder) from a main instance to a satellite instance, so
> for now I'd recommend just accessing your store as a client from your
> laptop.
>
>>
>> (2) Can the raspberry pi also be a local cache of my store on my network?
>
>
> I had an instance on a Beaglebone Black (~comparable to a Pi v2). It was
> fine for serving content, but too slow to keep up with chunking uploaded
> data and indexing. The Pi3 may be fast enough, but I haven't tried it yet.
>
> If the node isn't taking uploads directly, just syncing from another
> instance, it would only have to deal with indexing. It might also be
> possible to replicate the index from your primary node if you use an index
> backend with replication (mongo etc) but I haven't tried this.
>
>>
>> (3) Any tips for how my backup workflow should look like -- if I use
>> s3/GCS, should I still back up my laptop periodically on disks? If so,
>> what's the best way to back up the camlistore data?
>
>
> Are you planning to take periodic snapshots of your homedir using "camput
> file ~"? I don't use this functionality, so I can't advise. My impression is
> it's quite reliable.
>
> The backup method for Camlistore data depends on the storage backend. The
> different cloud storage providers all have their own best practices for
> backups. There are three different local filesystem backends (localdisk: one
> file per blob; diskpacked: 4GB journal files; and blobpacked, which stores
> blobs loose initially and then repacks for efficient sequential reads). I
> think for all of them you'd be safest to shut down the service during a
> backup. Modern filesystems like ZFS and BTRFS support online snapshots which
> is safest of all.
>
>>
>> (4) I think my $HOME should still be on disk, and not use the fuse mount;
>> in that case, how easy is it to keep it in sync with camlistore?
>
>
> There are a couple of ways to map a directory structure into Camlistore.
>
>  (a) camput file $PATH creates a static (deduplicated) snapshot of the path.
> It's great for backups, and is the foundation on which the other mode is
> built. You can restore with camget -o $OUTDIR $REF.
>  (b) camput file -filenodes $PATH creates a mutable "permanode" for each
> file/directory under the path. This mode is appropriate when you plan to
> treat Camlistore as the source of truth going forward.  Unlike (a), the
> command isn't idempotent; each run creates a new set of permanodes. You can
> also create permanodes by creating/copying directories under "roots/" in the
> FUSE mount.
>
> If you are talking about bidirectional sync between the local filesystem and
> permanodes, it's possible but requires some care. Always test first on a
> throwaway instance (devcam server -wipe is your friend). The main thing to
> be aware of: lots of user space programs spam temp files, and cammount
> doesn't filter them out (yet). Also, depending on how a sync program
> operates, it may remove the original file and create a new one. I'd like to
> document which sync software works, and flags are needed. As Mathieu said,
> you may want to wait. Help wanted :)

Sorry, what? Where did I say to wait? Wait for what?

-- 
You received this message because you are subscribed to the Google Groups 
"Camlistore" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to