After way too many days of debugging and diving into tech that looks like it was built in the year 2000, I finally found the culprit.

It wasn’t the code. It wasn’t the config. It was OS-level limitations. 😭😭😭


Seriously, that "think outside the box" advice is actually real.

I’ve been battling SIP stuff using FreeSwitch, and it kept failing mid-execution.

Since I’m running it on Docker, I didn't immediately suspect the OS itself.

After days of banging my head against the wall, I tweaked the kernel settings below and—boom—it actually works.

It still needs some optimization since it’s not a bare-metal server, but at least it’s stable now!

Here is the sysctl.conf magic that saved my sanity:

# etc/sysctl.conf

# Maximize the number of open files
fs.file-max = 2097152

# UDP Buffer Sizes - Essential for high-rate RTP/SIP packets (approx 256MB)
net.core.rmem_max = 268435456
net.core.wmem_max = 268435456
net.core.rmem_default = 268435456
net.core.wmem_default = 268435456

# Increase packets in the backlog queue
net.core.netdev_max_backlog = 100000

# Increase local ports available for UDP/TCP
net.ipv4.ip_local_port_range = 1024 65535

# Conntrack tuning - prevents the kernel from dropping high-traffic VoIP packets
net.netfilter.nf_conntrack_max = 1048576
net.netfilter.nf_conntrack_tcp_timeout_established = 3600

Victorious.

Let me get some rest then.

Good night