Closed smetj closed 10 years ago
I cannot reproduce that with the current stable release. However there was a problem with CLOSE_WAIT connection in previous releases, so i assume this is already fixed. I don't think the dns errors are related to Mod-Gearman or Gearman itself.
I based myself upon this statement in the documentation: "Latest stable version 1.4.2, released January 10 2013" https://labs.consol.de/nagios/mod-gearman/#_download
This piece of the documentation should be updated then as the latest stable is a newer version than indicated.
I'll upgrade to the latest version and evaluate
Thanks, i just updated the page to reflect latest updates.
Gearmand 0.25 Mod_gearman: 1.4.6 OS: rhel 6.2 Kernel 2.6.32-220.17.1.el6.x86_64
lsof command output
gearmand 16611 gearmand 9947u IPv4 3479410565 0t0 TCP master-101:gearman->worker-205:51992 (CLOSE_WAIT) gearmand 16611 gearmand 9950u IPv4 3479426713 0t0 TCP master-101:gearman->worker-102:42700 (CLOSE_WAIT) gearmand 16611 gearmand 9951u IPv4 3479365897 0t0 TCP master-101:gearman->worker-203:59654 (CLOSE_WAIT)
When all filehandles are consumed, the possibility to create a new connection to gearmand is sparse (until close_wait) times out.
It seems gearmand is not closing the socket when a client disconnects? This brings me to another topic, I wonder why the clients disconnect so often, ... we don't seem to have network issues yet, mod_gearman clients produce many of these: "[2013-09-01 22:36:38][31858][ERROR] sending job to gearmand failed: libgearman/connection.cc:482(GEARMAN_GETADDRINFO) Name or service not known -> getaddrinfo:%s"