View Issue Details

IDProjectCategoryView StatusLast Update
0000353bareos-coreGeneralpublic2016-01-14 13:52
Reporterhedererjs Assigned To 
PrioritylowSeverityfeatureReproducibilityalways
Status closedResolutionsuspended 
Summary0000353: Backup notebooks (and phones?)
DescriptionIn normal cases of bareos, director calls FD in order to start backup job.
But if FD is not callable (FD is outside enterprise network OR FD has DHCP and has changed IP address), feature request is :
 * FD calls DIR in order to create a socket :
   * FD configuration must include external IP address and external port for FD
   * FD calls DIR on schedule
 * DIR listens to FDs:
   * DIR configuration should include : MaxClientsPresence in order not to permit overload, Port for listening FDs (? or we use existing DIR port ?)
   * DIR opens sockets for listening FDs OR listens for new commands
 * when FD calls DIR, DIR looks at jobs waiting for FD and starts jobs
 * DIR automatically ends jobs that have waited for too much time

 
Additional InformationWe (ASPerience team) had proposed feature for bacula 2.4.4 but Kern said that this feature was not of interest for him. Like it was of interest for us, we developed it for bacula 2.4.4. Code is not very good when I read it today, but feature is good for me.
I'm trying to porting code to bareos master
What do you think of feature?
TagsNo tags attached.

Activities

mvwieringen

mvwieringen

2014-10-25 20:47

developer   ~0001023

I think I would like to make things a bit broader e.g. combine the ideas
with the feature request of allowing a client initiated backup.

You can probably re-use the socket that gets opened from the client to
the director after doing the normal challenge response mostly in the same
way as we do for passive backups in Bareos e.g. nothing prevents you from
using the opened socket as return socket mostly like it normally gets initiated
by the director.

I don't like the idea of using an extra listener and socket port. We could use
the same approach that the filed and stored already use in Bareos and that is
dispatch based on the hello message send.

Given you have very old code I wonder how much of it will still apply to
a code base that has drifted for almost 8 years. Bareos also is very different
from a config engine aspect etc.

If you can show the current code we can easily decide if it makes any sense
to reuse that code or to redesign things from scratch.

If we want to do a proper design first it probably makes sense to do that
via the devel mailinglist as this is a somewhat larger change we don't want
to end up with something thats unsupportable. As long as things are configurable
I see no problems in adding some changes that potential may be of use for some
people. At least people coming from TSM etc like the idea of client initiated
backups and even if it may not be for everyone I can see a use case there.

We probably first need to set the expectations right as we need to do proper
design first and then potentially reuse some already existing code.
mvwieringen

mvwieringen

2014-11-23 15:29

developer   ~0001074

I added to the master branch some refactoring to allow the DIR to determine
if its a normal console connection or a client connecting to the director.

We could take the design further from here. I would be interested in reviewing
some code even when its not working.

Issue History

Date Modified Username Field Change
2014-10-25 13:21 hedererjs New Issue
2014-10-25 20:47 mvwieringen Note Added: 0001023
2014-10-25 20:47 mvwieringen Status new => feedback
2014-11-23 15:23 mvwieringen Assigned To => mvwieringen
2014-11-23 15:23 mvwieringen Status feedback => confirmed
2014-11-23 15:29 mvwieringen Note Added: 0001074
2014-11-23 15:29 mvwieringen Status confirmed => feedback
2015-02-10 14:49 mvwieringen Assigned To mvwieringen =>
2016-01-14 13:52 mvwieringen Status feedback => closed
2016-01-14 13:52 mvwieringen Resolution open => suspended