On 08.01.21 19:04, Duarte Rocha wrote:
Thanks for the help.

I'm using db_mode 1 - Write-Through scheme. As far as i understand it uses database and cache correct? Is that the difference from mode 3?
Yes, db_mode 1 is different than db_mode 3, the later does no caching.

I'll do some testing with the usrloc's behaviour and i'll let you know.


If there is no external application blocking the database, look at latency core parameters, it may help to figure out if kamailio does a query that is slow.

Cheers,
Daniel


Cheers


Daniel-Constantin Mierla <miconda@gmail.com> escreveu no dia sexta, 8/01/2021 à(s) 17:47:

Hello,

if there is no external application that accesses the location table and you have many records, try to tune the behaviour of usrloc module, by setting timer procs and interval to different values. Also, if you use database-only mode with a lot of active users, you can also consider partitioning users on different tables, so you can have rules like users with id starting a-m go to location1 and n-z go to location2.

Cheers,
Daniel

On 08.01.21 18:29, Duarte Rocha wrote:
Hello Daniel, 

I have Kamailio on an HA setup. One machine has the active Kamailio and if something happens, the IP jumps to the another machine where there is a Kamailio running. Both Kamailio's share the same remote database. Could the reason be both Kamailio's trying to remove expired contacts from the location table at the same time? 

Cheers,
Duarte

Daniel-Constantin Mierla <miconda@gmail.com> escreveu no dia quinta, 7/01/2021 à(s) 19:28:

Hello,

is there any other application that accesses the location table?

That error message usually appears when the update of the record cannot be made because it is locked for long time by something else.

Cheers,
Daniel

On 07.01.21 18:23, Duarte Rocha wrote:
Hi all,

I've made a post last month regarding losing MySQL connections  - https://lists.kamailio.org/pipermail/sr-users/2020-December/111389.html

At the time I thought connections were dying as a consequence of low activity and traffic on the proxy. Meanwhile, I've migrated a great number of equipments to the proxy with Registers being refreshed every 10minutes and the problem still persists.

In order to try to fix this i've added timeout_interval and ping_interval from the db_mysql module. My SQL client on the Kamailio machine is mysql-community-client  5.6.50-2.el7. It writes and reads in a remote InnoDB database.

This are the logs i get from Kamailio when the problem appears :

Jan  7 09:43:27 sbc_bbt01_active /usr/local/kamailio-5.4/sbin/kamailio[21735]: ERROR: {1 27880 REGISTER e5f8f7bc-cbb2-40b3-9037-edacd6276a2b} db_mysql [km_dbase.c:123]: db_mysql_submit_query(): driver error on query: Lock wait timeout exceeded; try restarting transaction (1205)
Jan  7 09:43:27 sbc_bbt01_active /usr/local/kamailio-5.4/sbin/kamailio[21735]: ERROR: {1 27880 REGISTER e5f8f7bc-cbb2-40b3-9037-edacd6276a2b} <core> [db_query.c:348]: db_do_update(): error while submitting query
Jan  7 09:43:27 sbc_bbt01_active /usr/local/kamailio-5.4/sbin/kamailio[21735]: ERROR: {1 27880 REGISTER e5f8f7bc-cbb2-40b3-9037-edacd6276a2b} usrloc [ucontact.c:1147]: db_update_ucontact_ruid(): updating database failed
Jan  7 09:43:27 sbc_bbt01_active /usr/local/kamailio-5.4/sbin/kamailio[21735]: ERROR: {1 27880 REGISTER e5f8f7bc-cbb2-40b3-9037-edacd6276a2b} usrloc [ucontact.c:1663]: update_contact_db(): failed to update database
Jan  7 09:43:27 sbc_bbt01_active /usr/local/kamailio-5.4/sbin/kamailio[21735]: ERROR: {1 27880 REGISTER e5f8f7bc-cbb2-40b3-9037-edacd6276a2b} registrar [save.c:784]: update_contacts(): failed to update contact
Jan  7 09:43:27 sbc_bbt01_active /usr/local/kamailio-5.4/sbin/kamailio[21735]: ERROR: {1 27880 REGISTER e5f8f7bc-cbb2-40b3-9037-edacd6276a2b} sl [sl_funcs.c:414]: sl_reply_error(): stateless error reply used: I'm terribly sorry, server error occurred (1/SL)


Originally I had usrloc db_mode on mode 3 - DB-Only Scheme. In order to try to mitigate the issue I changed it to mode 1 - Write-Through scheme but even then I get the same log errors and an "500" error is still sent to the client. I chose this mode since, as far as I can understand it applies changes directly to DB but also uses cache. Please correct me if i'm wrong on that.

Has this issue happened with anyone before? Is there a way to mitigate this issue? My only constraint is that I need the database to be always updated since I have an HA setup, and as such, I can't use cache only methods.

Best Regards,


_______________________________________________
Kamailio (SER) - Users Mailing List
sr-users@lists.kamailio.org
https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users
-- 
Daniel-Constantin Mierla -- www.asipto.com
www.twitter.com/miconda -- www.linkedin.com/in/miconda
Funding: https://www.paypal.me/dcmierla
-- 
Daniel-Constantin Mierla -- www.asipto.com
www.twitter.com/miconda -- www.linkedin.com/in/miconda
Funding: https://www.paypal.me/dcmierla
-- 
Daniel-Constantin Mierla -- www.asipto.com
www.twitter.com/miconda -- www.linkedin.com/in/miconda
Funding: https://www.paypal.me/dcmierla