[Gluster-devel] Problems running AFR

[Matt Bennett]

Hello list,

Using 1.3.0-pre-2.3, I have configured a two server, single client
setup following [1]. Apart from the IP addresses, my config files are
identical to those in the user guide.

Both servers and client start and run happily. If I write to the
mounted filesystem with the client, the file I created appears in
/home/export on server1 and /home/afr-export on server2 (and visa
versa).

The problem comes when I take one of the servers down. If I take down
server1, the client can no longer write to the mounted filesystem - it
gives me the error "Transport endpoint is not connected", and the
client logs contain this:

[Apr 19 16:37:19] [ERROR/tcp-client.c:284/tcp_connect()]
tcp/client:non-blocking connect() returned: 111 (Connection refused)
[Apr 19 16:37:19] [ERROR/client-protocol.c:183/client_protocol_xfer()]
protocol/client: client_protocol_xfer: :transport_submit failed
[Apr 19 16:37:19]
[DEBUG/client-protocol.c:2543/client_protocol_cleanup()]
protocol/client:cleaning up state in transport object 0x8076f08
[Apr 19 16:37:19] [DEBUG/tcp-client.c:174/tcp_connect()] transport:
tcp: :try_connect: socket fd = 9
[Apr 19 16:37:19] [DEBUG/tcp-client.c:196/tcp_connect()] transport:
tcp: :try_connect: finalized on port `1021'
[Apr 19 16:37:19] [DEBUG/tcp-client.c:255/tcp_connect()]
tcp/client:connect on 9 in progress (non-blocking)
[Apr 19 16:37:19] [DEBUG/tcp-client.c:293/tcp_connect()]
tcp/client:connection on 9 still in progress - try later
[Apr 19 16:37:19] [ERROR/client-protocol.c:183/client_protocol_xfer()]
protocol/client: client_protocol_xfer: :transport_submit failed
[Apr 19 16:37:19]
[DEBUG/client-protocol.c:2543/client_protocol_cleanup()]
protocol/client:cleaning up state in transport object 0x8076f08
[Apr 19 16:37:19] [ERROR/tcp-client.c:284/tcp_connect()]
tcp/client:non-blocking connect() returned: 111 (Connection refused)
[Apr 19 16:37:19] [ERROR/client-protocol.c:183/client_protocol_xfer()]
protocol/client: client_protocol_xfer: :transport_submit failed
[Apr 19 16:37:19]
[DEBUG/client-protocol.c:2543/client_protocol_cleanup()]
protocol/client:cleaning up state in transport object 0x8076f08
[Apr 19 16:37:19] [DEBUG/tcp-client.c:174/tcp_connect()] transport:
tcp: :try_connect: socket fd = 9
[Apr 19 16:37:19] [DEBUG/tcp-client.c:196/tcp_connect()] transport:
tcp: :try_connect: finalized on port `1021'
[Apr 19 16:37:19] [DEBUG/tcp-client.c:255/tcp_connect()]
tcp/client:connect on 9 in progress (non-blocking)
[Apr 19 16:37:19] [DEBUG/tcp-client.c:293/tcp_connect()]
tcp/client:connection on 9 still in progress - try later
[Apr 19 16:37:19] [ERROR/client-protocol.c:183/client_protocol_xfer()]
protocol/client: client_protocol_xfer: :transport_submit failed
[Apr 19 16:37:19]
[DEBUG/client-protocol.c:2543/client_protocol_cleanup()]
protocol/client:cleaning up state in transport object 0x8076f08

If, however, I take down server2, then this problem doesn't occur.
Writes happen as normal, but obviously don't appear in
/home/afr-export on server1.

This brings me on to my next question - when does the resync happen
between servers? When I bring server2 back up again, files written to
server1 during the downtime are not copied to server2.

Is there a particular way to force the resync?

Many thanks,
Matt.














[1] http://www.gluster.org/docs/index.php/GlusterFS_User_Guide#AFR_Example_in_Clustered_Mode