[jitsi-dev] [jitsi-meet][jitsi-videobridge] SCTP connection not ready yet


#1

Hey,

I noticed that some participants cannot be connected. Doing a little
investigation I found that SCTP connection is not established for
them. Just disabling config.openSctp in meet solves this.

From JVB logs:

2015-05-01 21:22:41.691 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:22:45.465 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 8
2015-05-01 21:22:45.559 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:47.959 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:48.042 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:50.055 INFO: [82533] net.sf.fmj.media.Log.info()
Resetting queue, last seq added: 9223372036854775806, current seq:
24566
2015-05-01 21:22:50.450 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:50.584 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:52.994 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:53.125 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:54.890 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:22:55.530 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:55.662 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:56.391 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:22:57.891 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:22:58.070 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:58.090 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 8
2015-05-01 21:22:58.130 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:58.206 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:23:00.526 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:23:00.591 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:23:00.609 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:23:00.628 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:23:00.759 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:23:01.192 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.

And so on. Where 318 is the user id.

Just wondering if it can be solved somehow, otherwise it's easy to
determine Dominant Speaker Endpoint on client side from audio levels,
and do not use DataCannel at all.

Regards,
Zalmoxisus


#2

Hi,

···

On Mon, Jun 1, 2015 at 10:36 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Hey,

I noticed that some participants cannot be connected. Doing a little
investigation I found that SCTP connection is not established for
them. Just disabling config.openSctp in meet solves this.

Can you provide some more details about your deployment like JVB,
Meet, Jicofo versions you use ? Do you have more detailed info about
participants who can not be connected ?

Regards,
Pawel


#3

Hey Zalmoxisus,

···

On Mon, Jun 1, 2015 at 11:36 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Hey,

I noticed that some participants cannot be connected. Doing a little
investigation I found that SCTP connection is not established for
them. Just disabling config.openSctp in meet solves this.

From JVB logs:

2015-05-01 21:22:41.691 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:22:45.465 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 8
2015-05-01 21:22:45.559 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:47.959 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:48.042 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:50.055 INFO: [82533] net.sf.fmj.media.Log.info()
Resetting queue, last seq added: 9223372036854775806, current seq:
24566
2015-05-01 21:22:50.450 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:50.584 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:52.994 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:53.125 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:54.890 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:22:55.530 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:55.662 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:56.391 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:22:57.891 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:22:58.070 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:22:58.090 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 8
2015-05-01 21:22:58.130 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:22:58.206 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:23:00.526 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:23:00.591 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.
2015-05-01 21:23:00.609 INFO: [58399] net.sf.fmj.media.Log.info()
Shrinking packet queue to 15
2015-05-01 21:23:00.628 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:23:00.759 INFO: [58399] net.sf.fmj.media.Log.info()
Growing packet queue to 16
2015-05-01 21:23:01.192 WARNING: [58386]
org.jitsi.videobridge.Endpoint.warn() SCTP connection with 318 not
ready yet.

And so on. Where 318 is the user id.

Just wondering if it can be solved somehow, otherwise it's easy to
determine Dominant Speaker Endpoint on client side from audio levels,
and do not use DataCannel at all.

Well there's more to the DataChannel than just focusing on the active
speaker. You virtually can't do LastN without it. So we need to fix
this.

Could you please open a ticket?

Pawel could you please have a look?

Emil

--
https://jitsi.org


#4

Hi.

I use the latest versions. The differences with the default
configuration: use websockets (instead of bosh), user jids are
persistent, config.disableStats is true, startVideoMuted and
startAudioMuted are 1. And I use only audio (require access only to
the microphone).

It happens only for 0.01% of participants, and always to the same. I
didn't pay attention to it before as mostly the problems with
connection were related to the previous issue [1].

I weren't able to reproduce this by myself. Will try to get more
details from those users.

Regards,
Zalmoxisus

[1]: http://lists.jitsi.org/pipermail/dev/2015-May/024269.html

···

On Tue, Jun 2, 2015 at 10:44 AM, Paweł Domas <pawel.domas@jitsi.org> wrote:

Hi,

On Mon, Jun 1, 2015 at 10:36 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Hey,

I noticed that some participants cannot be connected. Doing a little
investigation I found that SCTP connection is not established for
them. Just disabling config.openSctp in meet solves this.

Can you provide some more details about your deployment like JVB,
Meet, Jicofo versions you use ? Do you have more detailed info about
participants who can not be connected ?

Regards,
Pawel

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


#5

Are you using bundle?

···

Am 02.06.2015 um 04:06 schrieb Michael Diordiev:

Hi.

I use the latest versions. The differences with the default
configuration: use websockets (instead of bosh), user jids are
persistent, config.disableStats is true, startVideoMuted and
startAudioMuted are 1. And I use only audio (require access only to
the microphone).

It happens only for 0.01% of participants, and always to the same. I
didn't pay attention to it before as mostly the problems with
connection were related to the previous issue [1].


#6

Hi,

···

On Tue, Jun 2, 2015 at 1:06 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Hi.

I use the latest versions. The differences with the default
configuration: use websockets (instead of bosh), user jids are
persistent, config.disableStats is true, startVideoMuted and
startAudioMuted are 1. And I use only audio (require access only to
the microphone).

It happens only for 0.01% of participants, and always to the same. I
didn't pay attention to it before as mostly the problems with
connection were related to the previous issue [1].

I weren't able to reproduce this by myself. Will try to get more
details from those users.

Does it happen to this 0,01% of participants every time or randomly ?

Does it work with Jicofo 94 ? The version with new endpoint ID
generation startegy, but with broken active speaker detection ?

Regarsd,
Pawel


#7

It happens randomly to those participants, but when it happens it
doesn't work even after reconnecting for those rooms. Now it works
smoothly even with openSCTP, but yesterday he had been tried for an
hour unsuccessfully till I disabled openSCTP.
He uses Chrome 42.0.2311.135 (Windows XP), and, as he reported, the
browser was working slowly as there were many participants.

There were problems with connection for Jicofo 94 as well.

Regards,
Zalmoxisus

···

On Tue, Jun 2, 2015 at 2:53 PM, Paweł Domas <pawel.domas@jitsi.org> wrote:

Hi,

On Tue, Jun 2, 2015 at 1:06 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Hi.

I use the latest versions. The differences with the default
configuration: use websockets (instead of bosh), user jids are
persistent, config.disableStats is true, startVideoMuted and
startAudioMuted are 1. And I use only audio (require access only to
the microphone).

It happens only for 0.01% of participants, and always to the same. I
didn't pay attention to it before as mostly the problems with
connection were related to the previous issue [1].

I weren't able to reproduce this by myself. Will try to get more
details from those users.

Does it happen to this 0,01% of participants every time or randomly ?

Does it work with Jicofo 94 ? The version with new endpoint ID
generation startegy, but with broken active speaker detection ?

Regarsd,
Pawel

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


#8

Yes, I do.

···

On Tue, Jun 2, 2015 at 3:33 PM, Philipp Hancke <fippo@goodadvice.pages.de> wrote:

Am 02.06.2015 um 04:06 schrieb Michael Diordiev:

Hi.

I use the latest versions. The differences with the default
configuration: use websockets (instead of bosh), user jids are
persistent, config.disableStats is true, startVideoMuted and
startAudioMuted are 1. And I use only audio (require access only to
the microphone).

It happens only for 0.01% of participants, and always to the same. I
didn't pay attention to it before as mostly the problems with
connection were related to the previous issue [1].

Are you using bundle?

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


#9

Hi Michael,

···

On Tue, Jun 2, 2015 at 1:06 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Hi.

I use the latest versions. The differences with the default
configuration: use websockets (instead of bosh), user jids are
persistent, config.disableStats is true, startVideoMuted and
startAudioMuted are 1. And I use only audio (require access only to
the microphone).

It happens only for 0.01% of participants, and always to the same. I
didn't pay attention to it before as mostly the problems with
connection were related to the previous issue [1].

The message "SCTP connection not ready yet" is displayed when
transport does not work, but I doubt that your issues are related to
SCTP channels. When the failure occurs do you see ICE failure on the
JVB (status changed to FAILED) ?

Regards,
Pawel


#10

Hi, Paweł.

I finally got access to one of the participants Teamviewer. He had
some malwares, that tried to show ads and somehow freezes jitsi-meet
on this point.
I was stuck in this as I didn't see other problems in logs, but it
doesn't seem to be an issue of JVB.
Sometimes I see other members that are able to connect only after few
retries, it also can be related to webrtc and slow computers.

I apologize for the invalid issue!

Regards,
Zalmoxisus

···

On Fri, Jun 5, 2015 at 12:15 PM, Paweł Domas <pawel.domas@jitsi.org> wrote:

Hi Michael,

On Tue, Jun 2, 2015 at 1:06 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Hi.

I use the latest versions. The differences with the default
configuration: use websockets (instead of bosh), user jids are
persistent, config.disableStats is true, startVideoMuted and
startAudioMuted are 1. And I use only audio (require access only to
the microphone).

It happens only for 0.01% of participants, and always to the same. I
didn't pay attention to it before as mostly the problems with
connection were related to the previous issue [1].

The message "SCTP connection not ready yet" is displayed when
transport does not work, but I doubt that your issues are related to
SCTP channels. When the failure occurs do you see ICE failure on the
JVB (status changed to FAILED) ?

Regards,
Pawel

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


#11

Hi Michael,

···

On Fri, Jun 5, 2015 at 12:29 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Hi, Paweł.

I finally got access to one of the participants Teamviewer. He had
some malwares, that tried to show ads and somehow freezes jitsi-meet
on this point.
I was stuck in this as I didn't see other problems in logs, but it
doesn't seem to be an issue of JVB.
Sometimes I see other members that are able to connect only after few
retries, it also can be related to webrtc and slow computers.

I apologize for the invalid issue!

No problem. Btw. the issue with persistent JIDs[1] is still valid,
after restoring old endpoint ID generation strategy. We're aware that
it may randomly fail and are working on fixing it.

[1]: https://github.com/jitsi/jicofo/issues/17

Regards,
Pawel


#12

Yes, it is the case, as Chrome had been crashed, the participant was
not able to enter the room till the session expires.

Regards,
Zalmoxisus

···

On Fri, Jun 5, 2015 at 1:48 PM, Paweł Domas <pawel.domas@jitsi.org> wrote:

Hi Michael,

On Fri, Jun 5, 2015 at 12:29 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Hi, Paweł.

I finally got access to one of the participants Teamviewer. He had
some malwares, that tried to show ads and somehow freezes jitsi-meet
on this point.
I was stuck in this as I didn't see other problems in logs, but it
doesn't seem to be an issue of JVB.
Sometimes I see other members that are able to connect only after few
retries, it also can be related to webrtc and slow computers.

I apologize for the invalid issue!

No problem. Btw. the issue with persistent JIDs[1] is still valid,
after restoring old endpoint ID generation strategy. We're aware that
it may randomly fail and are working on fixing it.

[1]: https://github.com/jitsi/jicofo/issues/17

Regards,
Pawel

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


#13

Hi Michael,

···

On Fri, Jun 5, 2015 at 1:00 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Yes, it is the case, as Chrome had been crashed, the participant was
not able to enter the room till the session expires.

In this case proably user is not able to join the MUC room, because
stalled connection is occupating the nickname. A possible solution to
this problem is adding a check for nickname conflict and trying
different one. This fix would have to be added to jitsi-meet.

Regards,
Pawel


#14

Thanks, Paweł. You pointed me in the right direction, and I solved this!

I didn't specify the resource identifier for the jid, so everytime
Prosody associates different resources. Having this one constant,
Prosody will destroy the last session automatically.

Actually, there were no problems with joining the MUC in this case,
only with JVB. So there's no need to have different nicknames. I use
userid as a nickname.

Regards,
Zalmoxisus

···

On Fri, Jun 5, 2015 at 2:09 PM, Paweł Domas <pawel.domas@jitsi.org> wrote:

Hi Michael,

On Fri, Jun 5, 2015 at 1:00 PM, Michael Diordiev <zalmoxisus@gmail.com> wrote:

Yes, it is the case, as Chrome had been crashed, the participant was
not able to enter the room till the session expires.

In this case proably user is not able to join the MUC room, because
stalled connection is occupating the nickname. A possible solution to
this problem is adding a check for nickname conflict and trying
different one. This fix would have to be added to jitsi-meet.

Regards,
Pawel

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