Hi Stefan,

did not looked to the code right now, but the approach by having the DMQ module doing the active/inactive handling centrally instead of duplicating it for all DMQ API users sounds better to me. We are also doing it in a similar way e.g. for specific database module errors.

Cheers,

Henning

From: Stefan Mititelu via sr-dev <sr-dev@lists.kamailio.org>
Sent: Montag, 17. März 2025 16:38
To: Kamailio (SER) - Development Mailing List <sr-dev@lists.kamailio.org>
Cc: Stefan Mititelu <stefan.mititelu@net2phone.com>
Subject: [sr-dev] DMQ improvements for detecting inactive nodes

Hi,

Currently there is ping mechanism in DMQ for detecting inactive nodes. However this triggers only on "ping_interval".

DMQ exposes API functions to other modules for sending dmq messages or broadcasting them. This API functions require "dmq_resp_cback_t" to be passed as parameter. Most of the other modules just log a DBG msg in it.


There are two improvement ideas:
1. other modules that use DMQ, to detect node failures too, as they are sending/bcasting messages. This, in addition to DMQ module "ping_interval" mechanism.

2. don't set inactive state of nodes on first failure, but have per node counter of fails, and modparam to tune this check.

To do this either:
A. Implement the response callback in each module that use DMQ, to check response code and set inactive state of dmq node => duplicate same code across different modules

B. Send NULL for that "dmq_resp_cback_t" from other modules that use DMQ. In DMQ module itself, check if NULL and if so, use a default callback that checks response code and set inactive state of dmq nodes => cleaner approach

Any opinions on this? Do you see any possible problems with this?

Thanks,

Stefan