Hey Rob,
thank you for the detailed reply. To be honest I had not thought about
VPN because of performance/throughput concerns but those are unwarranted
as my clients push to s3 via a storage daemon which has a public ip and
can be reached via a gateway and so the main traffic will not go through
the VPN.
For the start, with only a few setups the VPN solution could work, but I
see possible issues when there are more setups, as the ranges of the
local subnets of my setups do not have to be distinct and I don't see
how I could setup routing over VPNs when there are eg. two
192.168.0.0/24 subnets behind two different jumphosts and unfortunately
keeping those subnets distinct is not withing my reach.
Martin
On 15.12.23 18:41, Rob Gerber wrote:
Could you establish a site-to-site VPN link from your director's lan to
the remote lan that is currently only accessible from the jump host?
If you're concerned about the remote site having access to the central
lan with director on it, you could vlan tag all packets from remote lan
VPN and pass tagged traffic to director server, forbidding other clients.
If need be, maybe modify the idea so that the central director's server
has a site-to-site VPN link to the remote lan. Maybe more difficult to
do if the director doesn't have a public IP (so maybe the remote VPN
server will have difficulty reaching the director to complete the
tunnel?) Also, a network infrastructure link will be maintained on
something that isn't a piece of core network equipment (director
server), hiding the configuration from network admins.
MAYBE, you could give director access to remote lan via standard VPN
(one way, client initiated, road warrior, whichever term means "not site
to site VPN"). You could run into issues with the VPN connection
disconnecting. Maybe solve those issues by having a runbeforejob script
that verifies the tunnel is up, and if it isn't restarts the VPN
connection prior to the backup starting. However, if there's any
instance where the clients would need to reach out to the director, and
if the client initiated VPN proves to be unstable, you could have an
issue. I have no reason to believe that client initiated VPN is
unstable, but I guess it's possible. Also you would probably need to
initiate this connection entirely using command line tools, which I
haven't done but imagine is possible using openvpn or similar.
I'm sure there might be bacula features that cover these eventualities,
but I'm not a big enough bacula expert to know about them.
Robert Gerber
402-237-8692
r...@craeon.net <mailto:r...@craeon.net>
On Fri, Dec 15, 2023, 3:59 AM Martin Reissner <mreiss...@wavecon.de
<mailto:mreiss...@wavecon.de>> wrote:
Hello and sorry for the generic subject. My issue is as follows:
I have a centralized director which should be used to backup several
setups with multiple clients/fds in a cloud environment. In those
setups
there is only one gateway/jumphost with a public ip, the actual
clients/fds only have an address in an internal subnet and are
reachable
from outside via ssh-proxyjump from the gw/jumphost or via a
loadbalancer.
So far the only solutions I have come up with are portforwardings on
the
gw eg. port 19102 gets forwarded to client1 port 9102, 29102 to client2
9102 and so on. This works but is kind of tedious with many clients.
I read something about client initiated backups using the tray monitor.
I will look into that but scheduling backups on the clients/fds takes
away one of the main advantages of bacula, which is the centralized
scheduling.
Are there any further options that I might not have found or thought of?
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
<mailto:Bacula-users@lists.sourceforge.net>
https://lists.sourceforge.net/lists/listinfo/bacula-users
<https://lists.sourceforge.net/lists/listinfo/bacula-users>
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
