[jitsi-dev] Robust Jitsi deployment


#1

We're looking at leveraging jitsi videobridge in our environment, but are
wondering if there are some deployment schemes that we're not seeing in
order to provide things like failover, etc.

Most of the issues we're running into seem to center around prosody and how
the components are configured to know about one another (and the prosody
server). For example, jicofo and the bridge only attempt to connect to the
prosody server at startup--what if the prosody server went down? How would
the other components reconnect? Also, I know jicofo allows for being
configured with multiple bridges in its config, but is there any way to
have it discover new bridges on the fly? Or have new bridges spin up on
the fly without modifying the prosody configuration?

Wondering if the Jitsi team (or anyone) has deployed in such a way to allow
for failover to backup xmpp servers, jicofos and videobridges such that
there'd be minimal service interruption?

Thanks,
brian


#2

Hey Brian,

We don't currently have this natively so th best way to handle it right now
is to have multiple equivalen deployments

Emil

···

On Wednesday, April 8, 2015, Brian Baldino <brian@highfive.com> wrote:

We're looking at leveraging jitsi videobridge in our environment, but are
wondering if there are some deployment schemes that we're not seeing in
order to provide things like failover, etc.

Most of the issues we're running into seem to center around prosody and
how the components are configured to know about one another (and the
prosody server). For example, jicofo and the bridge only attempt to
connect to the prosody server at startup--what if the prosody server went
down? How would the other components reconnect? Also, I know jicofo
allows for being configured with multiple bridges in its config, but is
there any way to have it discover new bridges on the fly? Or have new
bridges spin up on the fly without modifying the prosody configuration?

Wondering if the Jitsi team (or anyone) has deployed in such a way to
allow for failover to backup xmpp servers, jicofos and videobridges such
that there'd be minimal service interruption?

Thanks,
brian

--
--sent from my mobile


#3

We're looking at leveraging jitsi videobridge in our environment, but are
wondering if there are some deployment schemes that we're not seeing in
order to provide things like failover, etc.

Most of the issues we're running into seem to center around prosody and how
the components are configured to know about one another (and the prosody
server). For example, jicofo and the bridge only attempt to connect to the
prosody server at startup--what if the prosody server went down? How would
the other components reconnect? Also, I know jicofo allows for being
configured with multiple bridges in its config, but is there any way to
have it discover new bridges on the fly? Or have new bridges spin up on
the fly without modifying the prosody configuration?

You can add components in prosody on the admin telnet console after doing a config:reload(). No restart required.

Wondering if the Jitsi team (or anyone) has deployed in such a way to allow
for failover to backup xmpp servers, jicofos and videobridges such that
there'd be minimal service interruption?

Failover is hard. It boils down to having the client go to a different set of servers usually. Which means if just your xmpp server is down you may have a number of JVBs that are unused.

Nothing prevents you from modifying the JVB in a way such that it can conect to multiple XMPP servers (I would strongly recommend not to do this under the same hostname though)

···

Am 08.04.2015 um 08:58 schrieb Brian Baldino:


#4

Thanks guys,
We're not XMPP experts but given the amount of large XMPP deployments, I
was wondering if there was something we weren't considering with regards to
deploying multiple XMPP servers to handle things like failover, as well as
any common practices for handling failover scenarios for jicofo and jvb.

Thanks again!
-brian

···

On Sat, Apr 11, 2015 at 1:43 PM, Philipp Hancke <fippo@goodadvice.pages.de> wrote:

Am 08.04.2015 um 08:58 schrieb Brian Baldino:

We're looking at leveraging jitsi videobridge in our environment, but are
wondering if there are some deployment schemes that we're not seeing in
order to provide things like failover, etc.

Most of the issues we're running into seem to center around prosody and
how
the components are configured to know about one another (and the prosody
server). For example, jicofo and the bridge only attempt to connect to
the
prosody server at startup--what if the prosody server went down? How
would
the other components reconnect? Also, I know jicofo allows for being
configured with multiple bridges in its config, but is there any way to
have it discover new bridges on the fly? Or have new bridges spin up on
the fly without modifying the prosody configuration?

You can add components in prosody on the admin telnet console after doing
a config:reload(). No restart required.

Wondering if the Jitsi team (or anyone) has deployed in such a way to

allow
for failover to backup xmpp servers, jicofos and videobridges such that
there'd be minimal service interruption?

Failover is hard. It boils down to having the client go to a different set
of servers usually. Which means if just your xmpp server is down you may
have a number of JVBs that are unused.

Nothing prevents you from modifying the JVB in a way such that it can
conect to multiple XMPP servers (I would strongly recommend not to do this
under the same hostname though)

_______________________________________________
dev mailing list
dev@jitsi.org
Unsubscribe instructions and other list options:
http://lists.jitsi.org/mailman/listinfo/dev