Description of problem: glusterfs process will listen port which is reserved in net.ipv4.ip_local_reserved_ports Version-Release number of selected component (if applicable): RHGS 3.2 How reproducible: every time Steps to Reproduce: 1. reserved ports by follow steps. # vi /etc/sysctl.conf Add the line at the bottom of the file: net.ipv4.ip_local_reserved_ports=700-900 Save and exit the file (:wq!) # sysctl -p # sysctl -a | grep ip_local_reserved_ports net.ipv4.ip_local_reserved_ports = 700-900 2. create a gluster volume and start it 3. gluster volume set VOL nfs.disable on 4. check the output of "netstat -anop | grep gluster" Actual results: In range 700-900, no port should be used. Expected results: glusterfs process will listen port by tcp and udp in the range. Additional info: A similar bug has been fixed in 2012 year https://bugzilla.redhat.com/show_bug.cgi?id=868715
These ports might be related to Gluster/NFS. It would be good to know what services are listening on those ports. Gluster/NFS requests a port from rpcbind (the portmapper) and uses that. In case rpcbind does not honour the net.ipv4.ip_local_reserved_ports sysctl option, Gluster is probably not responsible for it. Could you check the rpcbind registration with 'rpcbind -p $SERVER' and let us know if these ports have been registered with rpcbind?
(In reply to Niels de Vos from comment #7) > These ports might be related to Gluster/NFS. It would be good to know what > services are listening on those ports. Gluster/NFS requests a port from > rpcbind (the portmapper) and uses that. In case rpcbind does not honour the > net.ipv4.ip_local_reserved_ports sysctl option, Gluster is probably not > responsible for it. > > Could you check the rpcbind registration with 'rpcbind -p $SERVER' and let > us know if these ports have been registered with rpcbind? Hi, I didn't find a rpcbind with a "-p" option, do you mean "rpcinfo -p"? Here is the output. [root@node-01876647 ~]# sysctl -a | grep ip_local_res net.ipv4.ip_local_reserved_ports = 700-900 [root@node-01876647 ~]# netstat -anop | grep gluster tcp 0 0 0.0.0.0:756 0.0.0.0:* LISTEN 6932/glusterfs off (0.00/0/0) .... udp 0 0 0.0.0.0:754 0.0.0.0:* 6932/glusterfs off (0.00/0/0) .... [root@node-01876647 ~]# rpcinfo -p program vers proto port service .... 100021 1 udp 754 nlockmgr 100021 1 tcp 756 nlockmgr
The needinfo request[s] on this closed bug have been removed as they have been unresolved for 500 days