10 May
2006
10 May
'06
11:24 a.m.
Then, if system N fails, all systems >N also fail. That's not very redundant.
-----Original Message-----
From: Edson [mailto:4lists@gmail.com]
Sent: Wednesday, May 10, 2006 9:01 AM
To: Douglas Garstang; 'Bogdan-Andrei Iancu'; 'unplug'
Cc: users(a)openser.org
Subject: RE: [Users] clustering
Create a 'chain' of t_replicate (1->2->3->...->n). Than You
can have 'n'
Proxies in the 'cluster'...
Edson.
-----Original Message-----
From: users-bounces(a)openser.org
[mailto:users-bounces@openser.org] On
Behalf Of Douglas Garstang
Sent: quarta-feira, 10 de maio de 2006 11:49
To: Bogdan-Andrei Iancu; unplug
Cc: users(a)openser.org
Subject: RE: [Users] clustering
Yes, but t_replicate() only allows you to replicate to ONE
other OpenSER
system. Two OpenSER systems is not a redundant
configuration - 3 is
redundant.
-----Original Message-----
From: Bogdan-Andrei Iancu [mailto:bogdan@voice-system.ro]
Sent: Wednesday, May 10, 2006 5:28 AM
To: unplug
Cc: users(a)openser.org
Subject: Re: [Users] clustering
Hi,
try yo google on the mailing list - the topic was often
debated. search
also for t_replicate() and replication.
regards,
bogdan
unplug wrote:
_______________________________________________
Users mailing list
Users(a)openser.org
http://openser.org/cgi-bin/mailman/listinfo/users How to do the clustering of openser? Any
reference?
_______________________________________________
Users mailing list
Users(a)openser.org
http://openser.org/cgi-bin/mailman/listinfo/users
_______________________________________________
Users mailing list
Users(a)openser.org
http://openser.org/cgi-bin/mailman/listinfo/users
10 May
10 May
3:24 p.m.
New subject: [Users] clustering
yes, that's right - I will take a look to see how difficult is to change
t_replicate() to be able to sent to more destinations in parallel.
regards,
bogdan
Douglas Garstang wrote:
Then, if system N fails, all systems >N also fail.
That's not very redundant.
-----Original Message-----
From: Edson [mailto:4lists@gmail.com]
Sent: Wednesday, May 10, 2006 9:01 AM
To: Douglas Garstang; 'Bogdan-Andrei Iancu'; 'unplug'
Cc: users(a)openser.org
Subject: RE: [Users] clustering
Create a 'chain' of t_replicate (1->2->3->...->n). Than You
can have 'n'
Proxies in the 'cluster'...
Edson.
-----Original Message-----
From: users-bounces(a)openser.org
[mailto:users-bounces@openser.org] On
Behalf Of Douglas Garstang
Sent: quarta-feira, 10 de maio de 2006 11:49
To: Bogdan-Andrei Iancu; unplug
Cc: users(a)openser.org
Subject: RE: [Users] clustering
Yes, but t_replicate() only allows you to replicate to ONE
other OpenSER
>system. Two OpenSER systems is not a redundant configuration - 3 is
>redundant.
>
>
>
>>-----Original Message-----
>>From: Bogdan-Andrei Iancu [mailto:bogdan@voice-system.ro]
>>Sent: Wednesday, May 10, 2006 5:28 AM
>>To: unplug
>>Cc: users(a)openser.org
>>Subject: Re: [Users] clustering
>>
>>
>>Hi,
>>
>>try yo google on the mailing list - the topic was often
>>debated. search
>>also for t_replicate() and replication.
>>
>>regards,
>>bogdan
>>
>>unplug wrote:
>>
>>
>>
>>>How to do the clustering of openser? Any reference?
>>>
>>>_______________________________________________
>>>Users mailing list
>>>Users(a)openser.org
>>>http://openser.org/cgi-bin/mailman/listinfo/users
>>>
>>>
>>>
>>>
>>_______________________________________________
>>Users mailing list
>>Users(a)openser.org
>>http://openser.org/cgi-bin/mailman/listinfo/users
>>
>>
>>
>_______________________________________________
>Users mailing list
>Users(a)openser.org
>http://openser.org/cgi-bin/mailman/listinfo/users
>
> 
4 p.m.
New subject: [Users] clustering
Bogdan-Andrei Iancu wrote:
yes, that's right - I will take a look to see how
difficult is to change
t_replicate() to be able to sent to more destinations in parallel.
I'd really encourage everybody evaluating a clustered solution to try
cacheless usrloc in combination with mysql-cluster. Scales very well,
without the side-effect of inevitable usrloc inconsistency...
Just my 2 cents,
Andy
4:28 p.m.
New subject: [Users] clustering
I use this kind of clustering, but there is a performance penalty. Not so
high, but...
Also, in the meantime, how about try a double-chained t_replicate
configuration in conjunction with DB cache? In case of a double replication,
I understand that it would only update the DB. Am I thinking in a, at least,
reasonable way?
Edson.
-----Original Message-----
From: users-bounces(a)openser.org [mailto:users-bounces@openser.org] On
Behalf Of Andreas Granig
Sent: quarta-feira, 10 de maio de 2006 17:01
To: Bogdan-Andrei Iancu
Cc: users(a)openser.org
Subject: Re: [Users] clustering
Bogdan-Andrei Iancu wrote:
yes, that's right - I will take a look to see
how difficult is to change
t_replicate() to be able to sent to more destinations in parallel.
I'd really encourage everybody evaluating a clustered solution to try
cacheless usrloc in combination with mysql-cluster. Scales very well,
without the side-effect of inevitable usrloc inconsistency...
Just my 2 cents,
Andy
_______________________________________________
Users mailing list
Users(a)openser.org
http://openser.org/cgi-bin/mailman/listinfo/users
4:48 p.m.
New subject: [Users] clustering
Edson wrote:
I use this kind of clustering, but there is a
performance penalty. Not so
high, but...
Right, but most importantly it scales. If you observe performance
degradation, you can just add another host without needing to touch the
other instances. You just have to add the new instance to DNS-SRV or to
a load-balancer.
If performance becomes an issue in a t_replicate environment, adding
another host doesn't help you much, because each registration has to be
handled at every single host of the cluster. And registrations are
typically the most frequent requests in a system.
That's just beside the fact that if one host goes down, you will always
get an inconsistent usrloc cache among the hosts, because if the host
comes up again, you have to mysql-dump the location table from another
host and start the proxy. Between the moment of the dump and the moment
the proxy finishes reading the location table into the cache, you will
miss all registrations. This becomes a problem if you have a large
number of actively registering subscribers.
These two issues are the main reasons why I'm very happy with a
cache-less solution.
Andy
4:41 p.m.
New subject: [Users] clustering
On Wednesday 10 May 2006 04:00, Andreas Granig wrote:
Could you use a cached usrloc with a low enough update interval? Or does
OpenSER only write it's information to the database at that interval, not
read?
Also, cacheless usrloc is only available in the devel version, correct?
Thanks.
---Mike
Bogdan-Andrei Iancu wrote:
yes, that's right - I will take a look to see
how difficult is to change
t_replicate() to be able to sent to more destinations in parallel.
I'd really encourage everybody evaluating a clustered solution to try
cacheless usrloc in combination with mysql-cluster. Scales very well,
without the side-effect of inevitable usrloc inconsistency...
Just my 2 cents,
Andy
_______________________________________________
Users mailing list
Users(a)openser.org
http://openser.org/cgi-bin/mailman/listinfo/users
4:49 p.m.
New subject: [Users] clustering
Mike Williams wrote:
Could you use a cached usrloc with a low enough update
interval? Or does
OpenSER only write it's information to the database at that interval, not
read?
In caching modes, there's no possibility to re-read the cache from DB
(except on restart).
Also, cacheless usrloc is only available in the devel
version, correct?
Right.
Andy
6590
days inactive
6590
days old
6 comments
5 participants
participants (5)
-
Andreas Granig -
Bogdan-Andrei Iancu -
Douglas Garstang -
Edson -
Mike Williams