[Deploying Sakai] [sakai-nakamura] OAE Backup and Restore

[Ray Davis]

Pre-Sparse is pretty well covered, so post-Sparse is the question. 
Thanks for the summary view, Ian.

Snipping:

 > If you are on sparse then all your user content is in a DB and the
 > blobs are on filesystem, immutable. You can then do a normal backup as
 > you do with Sakai CLE. You should run Jackrabbit on a local Derby
 > instance and not back it up, since it should only ever be a reflection
 > of the contents of the bootstrap Jar.

I see your point, but in the short term I'll probably recommend getting 
as complete a view of the persisted state as possible (including the 
Jackrabbit storage). Even if there were no manually-triggered changes to 
Jackrabbit data after deployment, its state might help determine data 
upgrade procedures for newer versions of the server software.

Once OAE has a shared strategy for upgrades, this could change. In the 
meantime, the Jackrabbit backups should be fast and stable at least!

Best,
Ray

On 7/7/11 3:05 AM, Ian Boston wrote:
> It depends if you are pre-sparse or post sparse.
>
> If you are pre-sparse, all the data is in Jackrabbit and so you have
> to use Jackrabbit backup techniques. This involves backing up the DB
> and then trying to backup the local state which may be changing.
> Frankly, its not easy to do Hot and most people do it Cold. We run
> Jackrabbit in production to support portal state in Camtools 2008
> (Sakai 2.5), which uses recursive rsyncs to get a stable state of the
> local file system. Its worked for about 4 years now, but its quite
> scary and we have only had to do one recover, which IIRC was
> successful.
>
> If you are running pre-sparse in a cluster without buying a custom
> persistence manager form Day (Adobe), if they will sell it to you...
> then you will notice that the performance of the cluster is inversely
> proportional to the number of nodes in the cluster. Which is one of
> the reasons why I tried so hard to get the project to dedicate time to
> writing a new content system, eventually giving up and doing it in my
> own time. (bitter that I was forced to do that?, yes!, glad I did?,
> absolutely, :) ).
> In addition to the performance issue, most notable with updates, you
> have to watch the centralised Journal table which is the
> sychronization point over all nodes in the cluster (hence the perfect
> parallel slowdown), and ensure that when you use a backup, you replay
> that journal from the transaction the backup was created. Its vital to
> get the local state in sync with the journal otherwise the Lucene
> index inside Jackrabbit becomes out of sync.
>
> -------------------------------------------------------------------------
>
> If you are on sparse then all your user content is in a DB and the
> blobs are on filesystem, immutable. You can then do a normal backup as
> you do with Sakai CLE. You should run Jackrabbit on a local Derby
> instance and not back it up, since it should only ever be a reflection
> of the contents of the bootstrap Jar.
>
> The search index needs a bit more care, but its OOTB Solr so whatever
> technique works for you with Solr is good. AFAIK No one is running in
> a cluster with Solr yet and I believe there are some issues which I am
> trying to address at Cambridge. We will be running multi master for
> PostgreSQL and elected Master/Slave with replication for Solr to
> enable us to a) Scale elastically and b) do real time backup via
> replication.
>
>
>
> This message may bounce from production due to one of the lists not
> recognizing my subscription there.
>
> HTH
> Ian
>
> On 6 July 2011 17:33, Clay Fenlason<clay.fenlason at et.gatech.edu>  wrote:
>> I don't want to interrupt, but am copying the production list. It
>> seems the best channel for capturing discussion for production issues
>> as the OAE goes into wider use.
>>
>> ~Clay
>>
>> ---------- Forwarded message ----------
>> From: Ray Davis<ray at media.berkeley.edu>
>> Date: Wed, Jul 6, 2011 at 12:26 PM
>> Subject: [sakai-nakamura] OAE Backup and Restore
>> To: sakai-kernel<sakai-kernel at googlegroups.com>
>>
>>
>> How are NYU, CSU, Cambridge, and other OAE pilot schools planning to
>> back up and restore the application (particularly on a cluster)?
>>
>> I'm not finding any information post-Sparse, although there were some
>> useful notes on Jackrabbit-based Sakai 3 backups at
>>
>> http://groups.google.com/group/sakai-kernel/browse_thread/thread/bb58712ba4260982/
>> and
>> https://confluence.sakaiproject.org/display/KERNDOC/Your+Last+Day+with+Sakai+Nakamura...
>>
>> Looks like most of this should still apply but I might be missing
>> something important.
>>
>> Thanks,
>> Ray
>>
>> --
>> You received this message because you are subscribed to the Google
>> Groups "Sakai Nakamura" group.
>> To post to this group, send email to sakai-kernel at googlegroups.com.
>> To unsubscribe from this group, send email to
>> sakai-kernel+unsubscribe at googlegroups.com.
>> For more options, visit this group at
>> http://groups.google.com/group/sakai-kernel?hl=en.
>>
>> --
>> You received this message because you are subscribed to the Google Groups "Sakai Nakamura" group.
>> To post to this group, send email to sakai-kernel at googlegroups.com.
>> To unsubscribe from this group, send email to sakai-kernel+unsubscribe at googlegroups.com.
>> For more options, visit this group at http://groups.google.com/group/sakai-kernel?hl=en.
>>
>>
> _______________________________________________
> production mailing list
> production at collab.sakaiproject.org
> http://collab.sakaiproject.org/mailman/listinfo/production
>
> TO UNSUBSCRIBE: send email to production-unsubscribe at collab.sakaiproject.org with a subject of "unsubscribe"
>