[OpenSIPS-Users] HA ongoing call support in 2.4 - transaction replication

[Jeff Pyle]

Brett,

Yes, exactly.  Most of the components I'm familiar with are supposed to
live in the same rack with direct cabling between them for this kind of
thing.  People often separate them to some degree with various levels of
success, depending mostly upon the stability of the network between them.

In this case having the option would certainly be nice, and I would happily
accept the caveats.


- Jeff



On Thu, Feb 20, 2020 at 12:31 PM Brett Nemeroff <brett at nemeroff.com> wrote:

> Jeff, I think a lot of the problem is the active chit-chat in the updates.
> It’s a lot. I can see however your use case makes a lot of sense and for a
> redundant pair sitting next to each other, I think it *could* make sense,
> if it is fast enough.
>
> I’m curious if Razvan would consider making it an option that can be
> turned on with the understanding that it doesn’t work well across DC
> boundaries because of the timing and need to happen quickly for
> synchronization. There are plenty of infrastructure components with this
> guideline for similar reasons.
>
>
>
> On Thu, Feb 20, 2020 at 10:48 AM Jeff Pyle <jeff at ugnd.org> wrote:
>
>> Alexei,
>>
>> I see the article.  In summary, transactions are too complicated to
>> synchronize between nodes of a cluster because of their short timing
>> intervals and complex structures.  Instead the approach is to get the
>> messages of a transaction back to the individual node that owns the
>> transaction so it can process there.  Got it.
>>
>> For a cluster with many anycast nodes, this makes a lot of sense.  For a
>> simple active/standby setup, it prevents one from achieving a hitless
>> failover from one node to another if there are active transactions.
>> Bummer.  I'm sure Razvan and the team understood this when deciding on this
>> architecture.  Big picture their approach solves a lot more problems than
>> it creates, and it's very cool nonetheless.
>>
>>
>> - Jeff
>>
>>
>> On Thu, Feb 20, 2020 at 1:56 AM Alexey Vasilyev <
>> alexei.vasilyev at gmail.com> wrote:
>>
>>> Hi Jeff,
>>>
>>> Transactions are not replicated.
>>> Here
>>>
>>> https://blog.opensips.org/2018/03/21/full-anycast-support-in-opensips-2-4/
>>> Razvan explains why. Section "Distributed transactions handling".
>>>
>>>
>>>
>>> -----
>>> ---
>>> Alexey Vasilyev
>>> --
>>> Sent from:
>>> http://opensips-open-sip-server.1449251.n2.nabble.com/OpenSIPS-Users-f1449235.html
>>>
>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensips.org/pipermail/users/attachments/20200220/746aa1d6/attachment.html>