Red Hat Bugzilla – Bug 849291
[glusterfs-3.3.0q43]: clear-locks attempts to connect using privileged port
Last modified: 2015-11-03 18:04:37 EST
+++ This bug was initially created as a clone of Bug #825562 +++
clear-locks commands fail if rpc-auth-allow is not set as it attempts to connect via privileged port.
The CLI reports the error as:
Volume clear-locks unsuccessful
clear-locks getxattr command failed. Reason: Transport endpoint is not connected
### glusterd.vol.log has
[2012-05-27 12:54:33.217491] E [rpcsvc.c:491:rpcsvc_handle_rpc_call] 0-glusterd: Request received from non-privileged port. Failing request
--- Additional comment from firstname.lastname@example.org on 2012-05-28 02:49:01 EDT ---
Did it again. Sorry, it's qa43 not 34.
--- Additional comment from email@example.com on 2012-05-28 05:37:31 EDT ---
Created attachment 587189 [details]
--- Additional comment from firstname.lastname@example.org on 2012-05-28 06:02:06 EDT ---
Created attachment 587198 [details]
--- Additional comment from email@example.com on 2012-05-28 06:11:43 EDT ---
Created attachment 587200 [details]
--- Additional comment from firstname.lastname@example.org on 2012-05-28 06:15:34 EDT ---
Created attachment 587203 [details]
Each of these logs are for just the minutes that were in the clearlocks log. Let me know if that's not what you needed.
The work around exists to make this work even when insecure ports are used.
add below line to glusterd.vol:
" option rpc-auth.ports.insecure on"
Hence reducing the priority of the bug in RHS.
amar@supernova:~/work/glusterfs$ git diff
diff --git a/doc/glusterd.vol b/doc/glusterd.vol
index de17d8f..8268c3d 100644
@@ -4,5 +4,12 @@ volume management
option transport-type socket,rdma
option transport.socket.keepalive-time 10
option transport.socket.keepalive-interval 2
+ # enable below option only if there are not many ports free below 1024
+ # to bind() the client process. In this case, you would have to run :
+ # 'gluster volume set <VOL> server.allow-insecure yes'
+ # on all the volumes to the proper functioning.
+# option rpc-auth-allow-insecure yes
option transport.socket.read-fail-log off
The above patch is the only way at the moment to fix these things.
Also, Would like to close the bug with D(ocumentation)P(ending) flag set, once we have more better way of handling security (SSL/Kerberos etc), this issue will go away.
WORKSFORME always when we have the above two options set. WONTFIX the part of not having these options set for now.