Error if one of several gearmand servers is down

[adrianlzt]

Environment: RedHat 6.5 gearmand-server-0.33-2.x86_64 gearmand-0.33-2.x86_64 mod_gearman-1.5.2-1.el6.x86_64

mod_gearman_neb.conf

server=localhost:4730
server=172.16.1.31:4730

If server 172.16.1.31 is down, mod_gearman_neb writes this to log file:

[2015-02-27 10:43:51][2769][ERROR] sending job to gearmand failed: connect_poll(No route to host) getsockopt() failed -> libgearman/connection.cc:104
[2015-02-27 10:43:51][2769][ERROR] worker error: connect_poll(No route to host) getsockopt() failed -> libgearman/connection.cc:104
[2015-02-27 10:43:54][2769][ERROR] worker error: connect_poll(No route to host) getsockopt() failed -> libgearman/connection.cc:104

But, if 172.16.1.31 is up but gearmand-server down, all works properly.

I have compiled mod_gearman 1.5.2 to use gearmand-1.0.6 and works correctly. Packages: https://drive.google.com/open?id=0B84NO1oyhUXva3A5NHdnMThReGM&authuser=0 https://drive.google.com/open?id=0B84NO1oyhUXvWHNPMzhCcTE4UDQ&authuser=0 https://drive.google.com/open?id=0B84NO1oyhUXvMlNCN0IzQW9rOEU&authuser=0 https://drive.google.com/open?id=0B84NO1oyhUXvbUZtd1RnaDBXN0U&authuser=0

[adrianlzt]

The cure is worse than the disease. With gearmand-1.0.6, if no gearmand is available, mod_gearman_neb start to create sockets until file descriptors exhaustion. Also, with running gearmand, if Icinga tries to execute an active check it gets stuck.

Going back to gearmand 0.33 using pacemaker to give HA.

[sni]

i haven't made good results with any gearman version > 0.33. Here is a list of working versions: http://labs.consol.de/nagios/mod-gearman/#_supported_dependencies

[sni]

Could you give the latest release a try. I fixen an issue with segfaulting on connection errors.