[jitsi-dev] Troubleshooting and understanding the state of the jitsi-meet suite of services


#1

Hi,

I've had jitsi-meet running in a stable state for a number of months,
however I've been running into failure cases that I have not been able to
debug much -- such as no one being able to join any rooms, all working
except a particular room, or one participant not being able to enter.

The part that I'm finding difficult is that I can't narrow down a failure
as I haven't much visibility into the state of a specific conference and
it's participants from the perspectives of prosody, jicofo, and the
jitsi-videobridge.

The kind of things I think I want to see are:
- What users are in a particular room according to prosody
- Who the focus user thinks is there, and what it thinks everyone's state is
- Which connections and rooms on the video bridge correspond the ones used
by prosody and jicofo
- Enough associated data so that I can search the log files to find where
the fault may be

Am I correct in thinking that this sort of data is not currently exposed,
and that it would be helpful in debugging specific conferences when running
in a production scenario? Or is there a better way to go about it that I
haven't come across?

Thanks,
Simon


#2

I've found some helpful data is exposed when you enable logging to
InfluxDB, which is searchable using the jitsi/jiloin web interface.

Is the InfluxDB still a supported integration, or is it deprecated? It
doesn't work with the current supported 0.9 version, and there's an open
pull request to update the integration to work with the current supported
version (https://github.com/jitsi/jicoco/pull/1)

Unfortunately stats collection doesn't seem to be supported in the
lib-jitsi-meet branch, which is what I'm building off of at the moment. Is
there work being done on stats collection right now for lib-jitsi-meet, or
should I work on it with my team and submit a pull request?

Cheers,
Simon

···

On Wed, Dec 16, 2015 at 2:59 PM, Simon Ditner <spditner@gmail.com> wrote:

Hi,

I've had jitsi-meet running in a stable state for a number of months,
however I've been running into failure cases that I have not been able to
debug much -- such as no one being able to join any rooms, all working
except a particular room, or one participant not being able to enter.

The part that I'm finding difficult is that I can't narrow down a failure
as I haven't much visibility into the state of a specific conference and
it's participants from the perspectives of prosody, jicofo, and the
jitsi-videobridge.

The kind of things I think I want to see are:
- What users are in a particular room according to prosody
- Who the focus user thinks is there, and what it thinks everyone's state
is
- Which connections and rooms on the video bridge correspond the ones used
by prosody and jicofo
- Enough associated data so that I can search the log files to find where
the fault may be

Am I correct in thinking that this sort of data is not currently exposed,
and that it would be helpful in debugging specific conferences when running
in a production scenario? Or is there a better way to go about it that I
haven't come across?

Thanks,
Simon


#3

I've heard there is some way to connect to some of the services with a
groovy shell. Anyone happen to know how to go about doing that?

Cheers,
Simon

···

On Wed, Jan 20, 2016 at 10:52 AM Simon Ditner <spditner@gmail.com> wrote:

I've found some helpful data is exposed when you enable logging to
InfluxDB, which is searchable using the jitsi/jiloin web interface.

Is the InfluxDB still a supported integration, or is it deprecated? It
doesn't work with the current supported 0.9 version, and there's an open
pull request to update the integration to work with the current supported
version (https://github.com/jitsi/jicoco/pull/1)

Unfortunately stats collection doesn't seem to be supported in the
lib-jitsi-meet branch, which is what I'm building off of at the moment. Is
there work being done on stats collection right now for lib-jitsi-meet, or
should I work on it with my team and submit a pull request?

Cheers,
Simon

On Wed, Dec 16, 2015 at 2:59 PM, Simon Ditner <spditner@gmail.com> wrote:

Hi,

I've had jitsi-meet running in a stable state for a number of months,
however I've been running into failure cases that I have not been able to
debug much -- such as no one being able to join any rooms, all working
except a particular room, or one participant not being able to enter.

The part that I'm finding difficult is that I can't narrow down a failure
as I haven't much visibility into the state of a specific conference and
it's participants from the perspectives of prosody, jicofo, and the
jitsi-videobridge.

The kind of things I think I want to see are:
- What users are in a particular room according to prosody
- Who the focus user thinks is there, and what it thinks everyone's state
is
- Which connections and rooms on the video bridge correspond the ones
used by prosody and jicofo
- Enough associated data so that I can search the log files to find where
the fault may be

Am I correct in thinking that this sort of data is not currently exposed,
and that it would be helpful in debugging specific conferences when running
in a production scenario? Or is there a better way to go about it that I
haven't come across?

Thanks,
Simon


#4

Hi Simon,

You're right that it's possible to access all of the bridge internals with
a Groovy shell, we have some code that implements that, but it's not
integrated with the master branch and it's not actively maintained nor
supported.

That said, the code is very simple which makes it easy to unrot or to port
to a newer version of the bridge. We use it from time to time for debugging
and changing various parameters inside the bridge.

I can point you to the following URLs:

* https://github.com/jitsi/jitsi-videobridge-groovysh : This is a
standalone OSGi project that you can put in your libs/ directory and have
it loaded by the bridge as an OSGi service, but you need to know how to do
that.

* https://github.com/gpolitis/jitsi-videobridge/tree/groovysh : Is a more
recent implementation and has the groovy shell incorporated inside the
jitsi-videobridge code.

···

*
https://github.com/gpolitis/jitsi-videobridge/blob/groovysh/doc/groovysh.md
: Has some documentation on how to use the shell.

I hope this helps!

Best,
George

On Thu, Jan 21, 2016 at 11:00 AM, Simon Ditner <spditner@gmail.com> wrote:

I've heard there is some way to connect to some of the services with a
groovy shell. Anyone happen to know how to go about doing that?

Cheers,
Simon

On Wed, Jan 20, 2016 at 10:52 AM Simon Ditner <spditner@gmail.com> wrote:

I've found some helpful data is exposed when you enable logging to
InfluxDB, which is searchable using the jitsi/jiloin web interface.

Is the InfluxDB still a supported integration, or is it deprecated? It
doesn't work with the current supported 0.9 version, and there's an open
pull request to update the integration to work with the current supported
version (https://github.com/jitsi/jicoco/pull/1)

Unfortunately stats collection doesn't seem to be supported in the
lib-jitsi-meet branch, which is what I'm building off of at the moment. Is
there work being done on stats collection right now for lib-jitsi-meet, or
should I work on it with my team and submit a pull request?

Cheers,
Simon

On Wed, Dec 16, 2015 at 2:59 PM, Simon Ditner <spditner@gmail.com> wrote:

Hi,

I've had jitsi-meet running in a stable state for a number of months,
however I've been running into failure cases that I have not been able to
debug much -- such as no one being able to join any rooms, all working
except a particular room, or one participant not being able to enter.

The part that I'm finding difficult is that I can't narrow down a
failure as I haven't much visibility into the state of a specific
conference and it's participants from the perspectives of prosody, jicofo,
and the jitsi-videobridge.

The kind of things I think I want to see are:
- What users are in a particular room according to prosody
- Who the focus user thinks is there, and what it thinks everyone's
state is
- Which connections and rooms on the video bridge correspond the ones
used by prosody and jicofo
- Enough associated data so that I can search the log files to find
where the fault may be

Am I correct in thinking that this sort of data is not currently
exposed, and that it would be helpful in debugging specific conferences
when running in a production scenario? Or is there a better way to go about
it that I haven't come across?

Thanks,
Simon

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