OMAPI file descriptors leak on open/close TCP connections to 7911?

[Konstantin Orekhov]

Hello, folks!
I have recently started migration efforts for our DHCP infrastructure to CentOS7.2 and just ran into an issue with OMAPI in included into CentOS 4.2.5 version of ISC DHCP.
Unlike version 4.1.1 (stock version that came in CentOS6.7), I can kill OMAPI part of running dhcpd within minutes with this simple loop:
[root at spc02 conf]# while true; do echo | telnet 0 7911; ls -l /proc/`pgrep dhcpd`/fd | wc -l; done
Trying 0.0.0.0...
Connected to 0.
Escape character is '^]'.
Connection closed by foreign host.
21
Trying 0.0.0.0...
Connected to 0.
Escape character is '^]'.
Connection closed by foreign host.
22
Trying 0.0.0.0...
Connected to 0.
Escape character is '^]'.
Connection closed by foreign host.
23
Trying 0.0.0.0...
Connected to 0.
Escape character is '^]'.
Connection closed by foreign host.
24
Trying 0.0.0.0...
Connected to 0.
Escape character is '^]'.
Connection closed by foreign host.
25
<snip>
1025
Trying 0.0.0.0...
^C
Once the max number of open files is reached, OMAPI becomes unresponsive even though DHCP daemon appears to still server the requests, so have to restart dhcpd after that. This is causing an issue for us as ECV checks by the load-balancer are the ones that would kill OMAPI they are supposed to monitor :)

I also just complied and tried the same on 4.3.3 and got pretty much the same result with the only difference that dhcpd just dies completely instead of OMAPI becoming unresponsive.
As I mentioned, 4.1.1 does not have this issue and above while loop keeps a number of FDs at steady 10-15 range.
Anyone ran into this issue? Am I missing something in my config? Or this is a bug and strangely enough nobody ran into it? Ideas/suggestions?

Thanks! Konstantin Orekhov
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/dhcp-users/attachments/20160303/2dae26ba/attachment-0001.html>