[Gluster-users] Unresolved rdma question...

[Matthew Temple]

Hello, we're seeing lots of errors like this in the nfs.log file.   We have
a 4-brick distributed replicated volume.
I have noticed in discussion that there is an issue with the setting of the
rdma.listen-port, but I don't know where to change
that or if that's really the issue.   Beyond the discussion, I'd really
love to see an example by *someone who has actually fixed*
the problem, because I so often feel like I'm living in the strangely
split-brained Gluster world, where setup is very simple,
elegant, and clean, but when problems /do/ appear, it's difficult to know
what to do.   I saw a posting about how to
change the transport type by deleting and recreating the volume, but I'd
really more like to know what these log errors are telling me and why the
clients are refusing the connection.  Anyone have an answer?

Matt Temple


[2013-01-30 10:17:40.605867] W [rdma.c:4521:gf_rdma_handshake_pollerr]
(-->/usr/sbin/glusterfs(main+0x58a) [0x40741a]
(-->/usr/lib64/libglusterfs.so.0() [0x3333c3ed14]
(-->/usr/lib64/glusterfs/3.3.1/rpc-transport/rdma.so(+0x80f8)
[0x7fa4dd3160f8]))) 0-rpc-transport/rdma: gf2-client-1: peer ()
disconnected, cleaning up
[2013-01-30 10:17:40.608512] E [rdma.c:4604:tcp_connect_finish]
0-gf2-client-2: tcp connect to  failed (Connection refused)
[2013-01-30 10:17:40.608624] W [rdma.c:4187:gf_rdma_disconnect]
(-->/usr/sbin/glusterfs(main+0x58a) [0x40741a]
(-->/usr/lib64/libglusterfs.so.0() [0x3333c3ed14]
(-->/usr/lib64/glusterfs/3.3.1/rpc-transport/rdma.so(+0x8210)
[0x7fa4dd316210]))) 0-gf2-client-2: disconnect called (peer:)
[2013-01-30 10:17:40.608695] W [rdma.c:4521:gf_rdma_handshake_pollerr]
(-->/usr/sbin/glusterfs(main+0x58a) [0x40741a]
(-->/usr/lib64/libglusterfs.so.0() [0x3333c3ed14]
(-->/usr/lib64/glusterfs/3.3.1/rpc-transport/rdma.so(+0x80f8)
[0x7fa4dd3160f8]))) 0-rpc-transport/rdma: gf2-client-2: peer ()
disconnected, cleaning up
[2013-01-30 10:17:40.611499] E [rdma.c:4604:tcp_connect_finish]
0-gf2-client-3: tcp connect to  failed (Connection refused)
[2013-01-30 10:17:40.611612] W [rdma.c:4187:gf_rdma_disconnect]
(-->/usr/sbin/glusterfs(main+0x58a) [0x40741a]
(-->/usr/lib64/libglusterfs.so.0() [0x3333c3ed14]
(-->/usr/lib64/glusterfs/3.3.1/rpc-transport/rdma.so(+0x8210)
[0x7fa4dd316210]))) 0-gf2-client-3: disconnect called (peer:)
[2013-01-30 10:17:40.611682] W [rdma.c:4521:gf_rdma_handshake_pollerr]
(-->/usr/sbin/glusterfs(main+0x58a) [0x40741a]
(-->/usr/lib64/libglusterfs.so.0() [0x3333c3ed14]
(-->/usr/lib64/glusterfs/3.3.1/rpc-transport/rdma.so(+0x80f8)
[0x7fa4dd3160f8]))) 0-rpc-transport/rdma: gf2-client-3: peer ()
disconnected, cleaning up

------
Matt Temple
Director, Research Computing
Dana-Farber Cancer Institute.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130130/d51f6a42/attachment.html>