Trying to implement the database dumps feature in 1.2

0 views
Skip to first unread message

Dennis McEntire

unread,
Dec 14, 2009, 8:46:03 PM12/14/09
to XtreemFS

Hi!

I am trying to set up this automatic database dumps feature in the 1.2
version, but am running into some difficulties. First of all, I
couldn't really find documentation on how to set this up, so I am just
"shooting in the dark" a little on setting this up.

Correct me anywhere I need it:

Basically, it appears that I have to have the BabuDB service running
on both the DIR and MRC services.
I changed the workers to 10 (was zero)
I uncommented all the BabuDB directives in the dirconfig.properties
file and restarted the DIR daemon.

The errors that show up in the logs are:

RegisterServiceOperation DIR RqDisp | datestamp | BabuDBException: You
are not allowed to proceed this operation, because this DB is not
running in master-mode! (error code: NO_ACCESS)

I am not running SSL at the moment on any of the services.

Thanks for the help,

Dennis

Jan Stender

unread,
Dec 15, 2009, 8:06:45 AM12/15/09
to xtre...@googlegroups.com
Hi Dennis,

> I am trying to set up this automatic database dumps feature in the 1.2
> version, but am running into some difficulties. First of all, I
> couldn't really find documentation on how to set this up, so I am just
> "shooting in the dark" a little on setting this up.

Actually, database dumps are not a new feature. A dump is a version
independent XML representation of the entire MRC database, from which
the original database can be restored. We initially integrated the
feature in order to make it possible to migrate an MRC database if the
representation of data in the MRC is changed.

What we added with XtreemFS 1.2 is the ability to create these dumps in
the background, i.e. without blocking the MRC for client requests. Thus,
the feature can also be used to regularly create backups of the MRC
while the system is running and being accessed.

Dumps are not created automatically, though. You can trigger them
manually by using the 'xtfs_mrcdbtool' script, which is part of the
xtreemfs-tools package. There's a man page for the tool, as well as a
section the user guide.

> Correct me anywhere I need it:
>
> Basically, it appears that I have to have the BabuDB service running
> on both the DIR and MRC services.

Yes, BabuDB is embedded in the MRC and DIR, where it is used to store
all kinds of metadata. It's rather an embedded database that runs in the
same process as the server than a "service" in the classical sense, though.

> I changed the workers to 10 (was zero)

I suggest not to change this property value. Setting it to something
greater than 0 causes worker threads to be used for request processing
at database level, instead of using the thread from which the database
is accessed. So far, XtreemFS doesn't benefit from the use of worker
threads at database level.

> I uncommented all the BabuDB directives in the dirconfig.properties
> file and restarted the DIR daemon.
>
> The errors that show up in the logs are:
>
> RegisterServiceOperation DIR RqDisp | datestamp | BabuDBException: You
> are not allowed to proceed this operation, because this DB is not
> running in master-mode! (error code: NO_ACCESS)

It seems that you activated replication by uncommenting
replication-related properties. Since this is still an experimental
feature that hasn't yet been fully integrated in XtreemFS, I suggest not
to activate any of the "BabuDB replication configuration" properties.

Hope this helps and best regards,
Jan
Reply all
Reply to author
Forward
0 new messages