[SR-Users] kamailio crushes because of tcp errors

[Dmitry]

hel

>From time to time (it does not depend upon traffic) we see in logs:
 ERROR: <core> [tcp_read.c:274]: tcp_read_data(): error reading: Bad file descriptor (9) ([ip_address]:49832 -> [ip_address]5060) ERROR: <core> [tcp_read.c:1303]: tcp_read_req(): ERROR: tcp_read_req: error readingERROR: <core> [io_wait.h:608]: io_watch_del(): trying to delete already erased entry 3031 in the hash(0, 0, (nil)) flags 10)CRITICAL: <core> [tcp_read.c:1654]: handle_io(): io_watch_del failed for 0x7f8471c51820 id 867 fd 3031, state 0, flags 4018, main fd 179, refcnt 2 ([ip_address]:49832 -> [ip_address]:5060)
there are a lot of such TCP CRITICAL errors and the kamailio crushes
kamailio -vversion: kamailio 4.3.5 (x86_64/linux)flags: STATS: Off, USE_TCP, USE_TLS, TLS_HOOKS, USE_RAW_SOCKS, DISABLE_NAGLE, USE_MCAST, DNS_IP_HACK, SHM_MEM, SHM_MMAP, PKG_MALLOC, DBG_QM_MALLOC, USE_FUTEX, FAST_LOCK-ADAPTIVE_WAIT, USE_DNS_CACHE, USE_DNS_FAILOVER, USE_NAPTR, USE_DST_BLACKLIST, HAVE_RESOLV_RESADAPTIVE_WAIT_LOOPS=1024, MAX_RECV_BUFFER_SIZE 262144, MAX_LISTEN 16, MAX_URI_SIZE 1024, BUF_SIZE 65535, DEFAULT PKG_SIZE 8MBpoll method support: poll, epoll_lt, epoll_et, sigio_rt, select.id: unknowncompiled on 17:51:55 Mar  8 2016 with gcc 4.4.7
It is still not clear - what exactly causes such behaviour.
I want to recriate it in the lab - so what can cause it.?
May be a process erases the file descriptors of TCP connections?
Any help is appreciated.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sip-router.org/pipermail/sr-users/attachments/20160816/feff3319/attachment.html>