avahi-daemon stops working after undetermined period of time

[Brian McKee-2]

Hi All

First post - let me know if I'm doing something obviously stupid :-)

I'm running a mixed fixed IP LAN of Linux, Mac and Windows machines.

Some of the Linux boxes are running Ubuntu 8.04.1 Hardy.  I'm  
advertising VNC and sftp on those boxes with avahi.  It works fine  
for a while, then stops working.   The VNC advertised by the Macs  
always works.  Doing a 'service avahi-daemon restart' on the Ubuntu  
boxes brings it right back again.

I know 'it quits working sometimes' isn't much of a bug report :-)  
I'd like some suggestions on how I can better determine what's making  
it stop and when.

I did poke thru Trac and Launchpad briefly, and didn't really find  
anything that looked apropos of my issue.

Brian


_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Lennart Poettering-8]

On Tue, 23.12.08 09:46, Brian McKee (map@...) wrote:

> Hi All
>
> First post - let me know if I'm doing something obviously stupid :-)
>
> I'm running a mixed fixed IP LAN of Linux, Mac and Windows machines.
>
> Some of the Linux boxes are running Ubuntu 8.04.1 Hardy.  I'm  
> advertising VNC and sftp on those boxes with avahi.  It works fine for a
> while, then stops working.   The VNC advertised by the Macs always works.
>  Doing a 'service avahi-daemon restart' on the Ubuntu boxes brings it
> right back again.
>
> I know 'it quits working sometimes' isn't much of a bug report :-)  I'd
> like some suggestions on how I can better determine what's making it stop
> and when.

Have you checked syslog?

Lennart

--
Lennart Poettering                        Red Hat, Inc.
lennart [at] poettering [dot] net         ICQ# 11060553
http://0pointer.net/lennart/           GnuPG 0x1A015CC4
_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Brian McKee-2]

On 23-Dec-08, at 10:23 AM, Lennart Poettering wrote:

> On Tue, 23.12.08 09:46, Brian McKee (map@...) wrote:
>>
>> I know 'it quits working sometimes' isn't much of a bug  
>> report :-)  I'd
>> like some suggestions on how I can better determine what's making  
>> it stop
>> and when.
>
> Have you checked syslog?
>

Yes - sorry - should have mentioned.

Here's a typical series from this morning.  I checked and it wasn't  
working, so I did an avahi-daemon restart.
To me, it looks normal.

|  09:19:46 counter1 avahi-daemon[27213]: Got SIGTERM, quitting.
|  09:19:46 counter1 avahi-daemon[27213]: Leaving mDNS multicast  
group on interface eth0.IPv4 with address 192.168.1.114.
|  09:19:46 counter1 avahi-daemon[23554]: Found user 'avahi' (UID  
109) and group 'avahi' (GID 120).
|  09:19:46 counter1 avahi-daemon[23554]: Successfully dropped root  
privileges.
|  09:19:46 counter1 avahi-daemon[23554]: avahi-daemon 0.6.22  
starting up.
|  09:19:46 counter1 avahi-daemon[23554]: Successfully called chroot().
|  09:19:46 counter1 avahi-daemon[23554]: Successfully dropped  
remaining capabilities.
|  09:19:46 counter1 avahi-daemon[23554]: Loading service file /
services/sftp-ssh.service.
|  09:19:46 counter1 avahi-daemon[23554]: Loading service file /
services/ssh.service.
|  09:19:46 counter1 avahi-daemon[23554]: Loading service file /
services/vnc.service.
|  09:19:46 counter1 avahi-daemon[23554]: Joining mDNS multicast  
group on interface eth0.IPv4 with address 192.168.1.114.
|  09:19:46 counter1 avahi-daemon[23554]: New relevant interface  
eth0.IPv4 for mDNS.
|  09:19:46 counter1 avahi-daemon[23554]: Network interface  
enumeration completed.
|  09:19:46 counter1 avahi-daemon[23554]: Registering new address  
record for fe80::20b:cdff:fe67:7626 on eth0.*.
|  09:19:46 counter1 avahi-daemon[23554]: Registering new address  
record for 192.168.1.114 on eth0.IPv4.
|  09:19:46 counter1 avahi-daemon[23554]: Registering HINFO record  
with values 'I686'/'LINUX'.
|  09:19:47 counter1 avahi-daemon[23554]: Server startup complete.  
Host name is counter1.local. Local service cookie is 3562443840.
|  09:19:48 counter1 avahi-daemon[23554]: Service "counter1" (/
services/vnc.service) successfully established.
|  09:19:48 counter1 avahi-daemon[23554]: Service "counter1" (/
services/ssh.service) successfully established.
|  09:19:48 counter1 avahi-daemon[23554]: Service "counter1" (/
services/sftp-ssh.service) successfully established.

and as of now (10.30) it's not working again.  There's no more avahi
related entries in the log

Brian



_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Lennart Poettering-8]

On Tue, 23.12.08 10:35, Brian McKee (map@...) wrote:

>>> I know 'it quits working sometimes' isn't much of a bug report :-)  
>>> I'd
>>> like some suggestions on how I can better determine what's making it
>>> stop
>>> and when.
>>
>> Have you checked syslog?
> and as of now (10.30) it's not working again.  There's no more avahi
> related entries in the log

Hmm, could you please run "avahi-daemon --debug" in a terminal (as
root of course) and then let it run? As soon as Avahi
crashes/terminates you should then see some kind of message in that
terminal that explains what is going on.

Even better would be to run it inside gdb and then get a backtrace
when it crashes:

http://fedoraproject.org/wiki/StackTraces
https://wiki.ubuntu.com/Backtrace

Lennart

--
Lennart Poettering                        Red Hat, Inc.
lennart [at] poettering [dot] net         ICQ# 11060553
http://0pointer.net/lennart/           GnuPG 0x1A015CC4
_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Brian McKee-2]

On 23-Dec-08, at 10:46 AM, Lennart Poettering wrote:

> On Tue, 23.12.08 10:35, Brian McKee (map@...) wrote:
>
>>>> I know 'it quits working sometimes' isn't much of a bug report :-)
>>>> like some suggestions on how I can better determine what's  
>>>> making it
>>>> stop and when.
>>>
>>> Have you checked syslog?
>> and as of now (10.30) it's not working again.  There's no more avahi
>> related entries in the log
>
> Hmm, could you please run "avahi-daemon --debug" in a terminal (as
> root of course) and then let it run? As soon as Avahi
> crashes/terminates you should then see some kind of message in that
> terminal that explains what is going on.
>
> Even better would be to run it inside gdb and then get a backtrace
> when it crashes:

Well, it's stopped working again, but it hasn't crashed or exited.  When
I say 'stopped working' I mean the computer no longer shows up in either
Chicken of the VNC on OSX or Remote Desktop Client & nautilus in Ubuntu.
I ran it under debug this time.  Results below.  If I understand the
backtrace thing properly it will only produce useful information if the
daemon actually crashes - so doing that wouldn't be useful in this case?

It's still sitting there in this 'non working' state at the moment if
there's some other info I can check on?

Brian

| root@counter1:~# avahi-daemon --debug
| Found user 'avahi' (UID 109) and group 'avahi' (GID 120).
| Successfully dropped root privileges.
| chroot.c: chroot() helper started
| avahi-daemon 0.6.22 starting up.
| Successfully called chroot().
| Successfully dropped remaining capabilities.
| chroot.c: chroot() helper got command 02
| Loading service file /services/sftp-ssh.service.
| Loading service file /services/ssh.service.
| Loading service file /services/vnc.service.
| Joining mDNS multicast group on interface eth0.IPv4 with address  
192.168.1.114.
| New relevant interface eth0.IPv4 for mDNS.
| Network interface enumeration completed.
| Registering new address record for fe80::20b:cdff:fe67:7626 on eth0.*.
| Registering new address record for 192.168.1.114 on eth0.IPv4.
| Registering HINFO record with values 'I686'/'LINUX'.
| dbus-protocol.c: interface=org.freedesktop.Avahi.Server, path=/,  
member=GetAPIVersion
| dbus-protocol.c: interface=org.freedesktop.Avahi.Server, path=/,  
member=GetState
| Server startup complete. Host name is counter1.local. Local service  
cookie is 185304068.
| dbus-protocol.c: interface=org.freedesktop.Avahi.Server, path=/,  
member=EntryGroupNew
| dbus-entry-group.c: interface=org.freedesktop.Avahi.EntryGroup,  
path=/Client0/EntryGroup1, member=GetState
| dbus-entry-group.c: interface=org.freedesktop.Avahi.EntryGroup,  
path=/Client0/EntryGroup1, member=AddService
| dbus-entry-group.c: interface=org.freedesktop.Avahi.EntryGroup,  
path=/Client0/EntryGroup1, member=Commit
| Service "counter1" (/services/vnc.service) successfully established.
| Service "counter1" (/services/ssh.service) successfully established.
| Service "counter1" (/services/sftp-ssh.service) successfully  
established.



_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Lennart Poettering-8]

On Tue, 23.12.08 13:26, Brian McKee (map@...) wrote:

> Well, it's stopped working again, but it hasn't crashed or exited.  When
> I say 'stopped working' I mean the computer no longer shows up in either
> Chicken of the VNC on OSX or Remote Desktop Client & nautilus in Ubuntu.
> I ran it under debug this time.  Results below.  If I understand the
> backtrace thing properly it will only produce useful information if the
> daemon actually crashes - so doing that wouldn't be useful in this
> case?

Ah, I understood Avahi died due to some reason.

If Avahi services only show up initially but then go away this sounds
a lot as if you'd have some kind of firewall running or some hardware
issue that disallows mDNS multicast packets (udp port 5353) to be
received by your computer but allows them to be sent.

Please make sure your firewall is properly configured or
disabled. Also make sure your router properly supports
multicasting. Use wireshark to find out if packets properly travel in
both directions.

Lennart

--
Lennart Poettering                        Red Hat, Inc.
lennart [at] poettering [dot] net         ICQ# 11060553
http://0pointer.net/lennart/           GnuPG 0x1A015CC4
_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Brian McKee-2]

On 23-Dec-08, at 1:53 PM, Lennart Poettering wrote:
> If Avahi services only show up initially but then go away this sounds
> a lot as if you'd have some kind of firewall running or some hardware
> issue that disallows mDNS multicast packets (udp port 5353) to be
> received by your computer but allows them to be sent.

Ah - ok - that gives me something to chew on.  I'll report back.
Thanks for your help.

Brian



_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Brian McKee-2]

On 23-Dec-08, at 2:24 PM, Brian McKee wrote:

>
> On 23-Dec-08, at 1:53 PM, Lennart Poettering wrote:
>> If Avahi services only show up initially but then go away this sounds
>> a lot as if you'd have some kind of firewall running or some hardware
>> issue that disallows mDNS multicast packets (udp port 5353) to be
>> received by your computer but allows them to be sent.
>
> Ah - ok - that gives me something to chew on.  I'll report back.

Yep - I bet that was it - I haven't left it running long enough to  
prove for sure, but the firewall wasn't opened correctly.  Thought  
I'd done that earlier :-(

Do I need the entire 224.0.0.0/4 open?  or can I tighten that down a  
bit?  Just 224.0.0.251 didn't seem to be enough.

Brian


_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Brian McKee-2]

On 23-Dec-08, at 2:32 PM, Brian McKee wrote:

> On 23-Dec-08, at 2:24 PM, Brian McKee wrote:
>
>>
>> On 23-Dec-08, at 1:53 PM, Lennart Poettering wrote:
>>> If Avahi services only show up initially but then go away this  
>>> sounds
>>> a lot as if you'd have some kind of firewall running or some  
>>> hardware
>>> issue that disallows mDNS multicast packets (udp port 5353) to be
>>> received by your computer but allows them to be sent.
>>
>> Ah - ok - that gives me something to chew on.  I'll report back.
>
> Yep - I bet that was it - I haven't left it running long enough to  
> prove for sure, but the firewall wasn't opened correctly.  Thought  
> I'd done that earlier :-(
>
> Do I need the entire 224.0.0.0/4 open?  or can I tighten that down  
> a bit?  Just 224.0.0.251 didn't seem to be enough.

Sorry to keep adding to my own thread here, but that hasn't gotten it  
either.  I'll drop the firewall altogether for a bit and see.

Brian



_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Lennart Poettering-8]

On Tue, 23.12.08 14:32, Brian McKee (map@...) wrote:

> On 23-Dec-08, at 2:24 PM, Brian McKee wrote:
>
>>
>> On 23-Dec-08, at 1:53 PM, Lennart Poettering wrote:
>>> If Avahi services only show up initially but then go away this sounds
>>> a lot as if you'd have some kind of firewall running or some hardware
>>> issue that disallows mDNS multicast packets (udp port 5353) to be
>>> received by your computer but allows them to be sent.
>>
>> Ah - ok - that gives me something to chew on.  I'll report back.
>
> Yep - I bet that was it - I haven't left it running long enough to prove
> for sure, but the firewall wasn't opened correctly.  Thought I'd done
> that earlier :-(
>
> Do I need the entire 224.0.0.0/4 open?  or can I tighten that down a  
> bit?  Just 224.0.0.251 didn't seem to be enough.

You need to allow mDNS traffic to port 5353 coming from any host on
the LAN and going to any host on the local LAN or going to the mDNS
group 224.0.0.251.

Lennart

--
Lennart Poettering                        Red Hat, Inc.
lennart [at] poettering [dot] net         ICQ# 11060553
http://0pointer.net/lennart/           GnuPG 0x1A015CC4
_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Brian McKee-2]

On 23-Dec-08, at 4:25 PM, Lennart Poettering wrote:

> On Tue, 23.12.08 14:32, Brian McKee (map@...) wrote:
>
>> On 23-Dec-08, at 2:24 PM, Brian McKee wrote:
>>
>>>
>>> On 23-Dec-08, at 1:53 PM, Lennart Poettering wrote:
>>>> If Avahi services only show up initially but then go away this  
>>>> sounds
>>>> a lot as if you'd have some kind of firewall running or some  
>>>> hardware
>>>> issue that disallows mDNS multicast packets (udp port 5353) to be
>>>> received by your computer but allows them to be sent.
>>>
>>> Ah - ok - that gives me something to chew on.  I'll report back.
>>
>> Yep - I bet that was it - I haven't left it running long enough to  
>> prove
>> for sure, but the firewall wasn't opened correctly.  Thought I'd done
>> that earlier :-(
>>
>> Do I need the entire 224.0.0.0/4 open?  or can I tighten that down a
>> bit?  Just 224.0.0.251 didn't seem to be enough.
>
> You need to allow mDNS traffic to port 5353 coming from any host on
> the LAN and going to any host on the local LAN or going to the mDNS
> group 224.0.0.251.

Just as a partial update to this issue.  The problem seems to be on  
the Mac side, not Avahi.  Have dropped for now.

Brian


_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Bryon Roche]

On Tue, 23 Dec 2008 22:25:36 +0100, Lennart Poettering wrote:

> LAN and going to any host on the local LAN or going to the mDNS group
> 224.0.0.251.

This will also hit you if you are using a default Fedora 10 install.

/B

_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Brian McKee-2]

I'm resurrecting an old thread here.

I have a mixed Mac / Linux LAN - all static IP.
I'm advertising a couple of services on the Linux boxes.
The Macs are obviously advertising themselves as well.

Each time avahi-daemon is restarted service discovery etc. all seems  
to work fine.
After about 10 minutes, you can't resolve the host names of the Linux  
boxes anymore.

Previously on this list (Dec 08) it was suggested I check for firewall  
issues - I've done so and confirmed that isn't currently a problem.

http://ubuntuforums.org/showthread.php?t=872827 shows I'm not alone  
anyway...

Suggestions on how to track down what's happening here?

I have run the avahi-daemon in --debug mode, and there is no  
indication there at all when resolution starts failing.

Brian
_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Brian McKee-2]

On 11-Sep-09, at 11:56 AM, Brian McKee wrote:

> I'm resurrecting an old thread here.
>
> I have a mixed Mac / Linux LAN - all static IP.
> I'm advertising a couple of services on the Linux boxes.
> The Macs are obviously advertising themselves as well.
>
> Each time avahi-daemon is restarted service discovery etc. all seems  
> to work fine.
> After about 10 minutes, you can't resolve the host names of the  
> Linux boxes anymore.

What's a good way to test for switch issues?  Can I use netcat or ? to  
prove that out one way or the other?

Brian
_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Trent Lloyd]

Brian,

On 12/09/2009, at 1:26 AM, Brian McKee wrote:

> On 11-Sep-09, at 11:56 AM, Brian McKee wrote:
>
>> I'm resurrecting an old thread here.
>>
>> I have a mixed Mac / Linux LAN - all static IP.
>> I'm advertising a couple of services on the Linux boxes.
>> The Macs are obviously advertising themselves as well.
>>
>> Each time avahi-daemon is restarted service discovery etc. all seems
>> to work fine.
>> After about 10 minutes, you can't resolve the host names of the
>> Linux boxes anymore.
>
> What's a good way to test for switch issues?  Can I use netcat or ? to
> prove that out one way or the other?

This whole issue is a tell-tale sign of a broken network device.

Usually it is wireless network cards, but sometimes is also wired and  
switches.

The way to tell is to run tcpdump on a "working" system, if you run a  
query on the originating host and don't see it on the network - then  
it is broken in some way.

But it seems you are describing a slightly reverse issue:

You restart Avahi-daemon on *this* system and it can resolve hostnames  
for 10 minutes and then stops?



The usual issue is you restart Avahi on a particular system and all  
systems can now see that node for 10 minutes, after that time the TTLs  
expire and any queries from other devices for its names or services  
are not seen as the network driver doesn't send the multicast requests  
to Avahi.

Trent
_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi

[Brian McKee-2]

On 17-Sep-09, at 1:07 AM, Trent Lloyd wrote:
> This whole issue is a tell-tale sign of a broken network device.
>
> Usually it is wireless network cards, but sometimes is also wired and
> switches.

It happens on every system.  They are all wired but various hardware.  
I can't believe that all those various network cards have driver issues.
Most of them are running Ubuntu 8.04LTS, but the one time I tested  
9.04 it showed similar behaviour.

If it was the switches, which are homogenous, then why would the Macs  
still see Bonjour advertised services?
Or is their software stack different enough that's a red herring?
The Macs always see each other, they only see a linux boxes after that  
box has it's avahi-daemon restarted.


> The usual issue is you restart Avahi on a particular system and all
> systems can now see that node for 10 minutes, after that time the TTLs
> expire and any queries from other devices for its names or services
> are not seen as the network driver doesn't send the multicast requests
> to Avahi.

Yes, that is what is happening. Apologies if the previous explanation  
was unclear.


I'll pull out Wireshark when I get a minute and see what I can see.  I  
wish I had a working network to contrast it to.

Brian


_______________________________________________
avahi mailing list
avahi@...
http://lists.freedesktop.org/mailman/listinfo/avahi