On 10-11 11:41, Bogdan-Andrei IANCU wrote:
Jan Janak wrote:
On 10-11 10:01, Bogdan-Andrei IANCU wrote:
my first guess is that use at start-up time you have an intensive private memory consumer which gets into conflict with the SIP messages receiver.
Judging after the error (in receive loop) I'm sure you have some SIP traffic at start-up - please check with tcpdump or ngrep - it can be a REGISTER, SUBSCRIBE, etc
Looking into your script, I would say only usrloc can be the intensive private memory consumer - if you have a lot of entries in DB (usrloc and aliases). To check this theory, try to start ser with DB disabled in usrloc; set: modparam("usrloc", "db_mode", 0)
please see if you still get the error and also please confirm if you have or not sip traffic.
I don't think this is the problem. usrloc consumes private memory before SER forks (and at that stage it does not process SIP messages), when re-loading data from the database. It would fail to start if this was the problem, so I think there must be some other memory problem.
I was considering the possibility of memory fragmentation. If there are a lot of usrloc/aliases records to be loaded, memory can get to fragmented to be able later to alloced a bigger size chunk.
Bot tables (aliases, location) are empty:
0(0) preload_udomain(): Table is empty 0(0) fixing /usr/local/lib/ser/modules/sl.so sl_send_reply 0(0) fixing /usr/local/lib/ser/modules/registrar.so lookup 0(0) preload_udomain(): Table is empty
Jan.