Backup strategy

[François Cassistat <f....@maya-systems.com>]

Hi,

I need to make daily backups of our repository on the server. But there may be user activity at any hour.

I was thinking about a backup strategy :
- Hold any new connection/request ;
- Let all connected users finish their current request ;
- Ensure that all connections are closed (with transient repository, it should shut down itself) ;
- Copy all database files of the persistence manager (no datastore, no indexes) directly by the filesystem ;
- Reactivate repository and performs new requests ;
- Begin copying files from the datastore. This may be incremental.
Maybe I should need to prevent files from disappearing from the datastore by moving them to a temporary trash node.

I would like to know what you think.


F

Re: Backup strategy

[Thomas Mueller <mu...@adobe.com>]

Hi,

Sorry for the delay.

>I would like to know how files in file data store are going to react in
>these cases :
>- If a user deletes a document during the data store backup process, do I
>need to move the file node to a temporary trash node make sure it is not
>removed from the data store?

Only the data store garbage collection removed files from the data store.

>- If a user adds a document during the data store backup process, may it
>create problems when we try to recover from a backup?

It depends on when you backup the data store. An easy solution is to
backup the data store after backing up everything else. Because only the
data store garbage collection removed files from the data store, this is
safe.

Regards,
Thomas

Re: Backup strategy

[François Cassistat <f....@maya-systems.com>]

I found that H2 gives me possibility to make a hot backups, so no need to block requests then. I'll give it a try.

I checked http://wiki.apache.org/jackrabbit/BackupAndMigration
and I would like to know how files in file data store are going to react in these cases :
- If a user deletes a document during the data store backup process, do I need to move the file node to a temporary trash node make sure it is not removed from the data store?
- If a user adds a document during the data store backup process, may it create problems when we try to recover from a backup?


F


Le 2011-02-04 à 5:22 PM, François Cassistat a écrit :

> Hi,
> 
> I need to make daily backups of our repository on the server. But there may be user activity at any hour.
> 
> I was thinking about a backup strategy :
> - Hold any new connection/request ;
> - Let all connected users finish their current request ;
> - Ensure that all connections are closed (with transient repository, it should shut down itself) ;
> - Copy all database files of the persistence manager (no datastore, no indexes) directly by the filesystem ;
> - Reactivate repository and performs new requests ;
> - Begin copying files from the datastore. This may be incremental.
> Maybe I should need to prevent files from disappearing from the datastore by moving them to a temporary trash node.
> 
> I would like to know what you think.
> 
> 
> F
>