Hi Murali,

no this didn't help.   (In my versions (cvs and 1.4.0) I didn't see a
src/io/bmi/bmi-tcp.c and  I made the chance  in
src/io/bmi/bmi_tcp/bmi-tcp.c on  line 2050).
--------8<---------------
[EMAIL PROTECTED] bmi_tcp]# /usr/local/pvfs2_nodes/bin/pvfs2-cp sockio.o
/mnt/pvfs2/
[E 10:07:29.224827] Receive immediately failed: Value too large for
defined data type
[E 10:07:29.224933] msgpair failed, will retry:: Value too large for
defined data type
[E 10:07:29.224970] *** msgpairarray_completion_fn: msgpair to server
tcp://node10:3334 failed: Value too large for defined data type
[E 10:07:29.224981] *** Non-BMI failure.
[E 10:07:29.224991] getattr_object_getattr_failure : Value too large for
defined data type
PVFS_sys_create: Value too large for defined data type
Could not open /mnt/pvfs2/
Segmentation fault
-------->8---------------



BTW,  after I set up a 1.8TB pvfs2,  using the kernel module and "cp", I
get a segfault copying within the pvfs2 filesystem.   According to
"diff" the file copy itself succeeded. 

... Matt


Murali Vilayannur wrote:

>Hi Matt,
>Could you replace line 2083 in src/io/bmi/bmi-tcp.c from
>int copy_size = 0;
>to    bmi_size_t copy_size = 0;
>and see if that helps?
>Thanks,
>Murali
>
>On Fri, 26 May 2006, Matt wrote:
>
>  
>
>>Hi
>>
>>I did some more testing.
>>
>>My problems may  to be related to a 2TB limit.  If my combined storage size
>>is below 2TB (9 nodes  a 200GB), everything seems to work.  However, if
>>I use 10 or more nodes I get the reported error.
>>
>>The same holds if I stay away from our head node and use only Mandriva
>>nodes with
>>FC4 kernel.  However, in this case the error message changes:
>>
>>---------8<---------------
>>[EMAIL PROTECTED] pvfs2]# /usr/local/pvfs2_nodes/bin/pvfs2-cp -t
>>/home/munnich/Soft/pvfs2/pvfs2-1.4.0.tar.gz /mnt/pvfs2/testfile
>>[E 14:22:06.071020] Receive immediately failed: Value too large for
>>defined data type
>>[E 14:22:06.071114] msgpair failed, will retry:: Value too large for
>>defined data type
>>[E 14:22:06.071159] *** msgpairarray_completion_fn: msgpair to server
>>tcp://node10:3334 failed: Value too large for defined data type
>>[E 14:22:06.071170] *** Non-BMI failure.
>>[E 14:22:06.071179] getattr_object_getattr_failure : Value too large for
>>defined data type
>>PVFS_sys_create: Value too large for defined data type
>>Could not open /mnt/pvfs2/testfile
>>Segmentation fault (core dumped)
>>--------->8---------------
>>
>>Does this behavior ring a bell?
>>
>>... Matt
>>
>>
>>Log files
>>[EMAIL PROTECTED] pvfs2]# cat /tmp/pvfs2-server.log
>>[D 14:21:59.274612] PVFS2 Server version 1.4.1pre1-2006-05-25-230553
>>starting.
>>[D 14:21:59.275359] Passing tcp://node10:3334 as BMI listen address.
>>[D 14:21:59.275417] BMI_tcp_initialize: Initializing TCP/IP module.
>>[D 14:21:59.275495] BMI_tcp_initialize: TCP/IP module successfully
>>initialized.
>>[D 14:21:59.276813] dbpf_thread_initialize: initialized
>>[D 14:21:59.278074] collection lookup: version is 0.1.2
>>[D 14:21:59.278238] dbpf_thread_function started
>>[D 14:21:59.278312] - set handle re-use timeout to 360 seconds (ret=0)
>>[D 14:21:59.301219] File system pvfs2-fs using handles:
>>        4-390451575
>>[D 14:21:59.301276] Sync on metadata update for pvfs2-fs: yes
>>[D 14:21:59.301287] Sync on I/O data update for pvfs2-fs: no
>>[D 14:21:59.301320] Storage Init Complete (aio-threaded)
>>[D 14:21:59.301331] 1 filesystem(s) initialized
>>[D 14:21:59.301816] Initialization completed successfully.
>>[D 14:22:06.068882] handle_new_connection: Assigning socket 12 to new
>>method addr.
>>[D 14:22:06.068956] tcp_do_work_recv: Reading header for new op.
>>[D 14:22:06.068972] tcp_do_work_recv: Received new message; mode: 2.
>>[D 14:22:06.068983] tcp_do_work_recv: tag: 1
>>[D 14:22:06.069054] (0x5e1b70) getconfig (prelude sm) state: req_sched
>>[D 14:22:06.069118] (0x5e1b70) getconfig (prelude sm) state:
>>getattr_if_needed
>>[D 14:22:06.069132] (0x5e1b70) getconfig (prelude sm) state: perm_check
>>(status = 0)
>>[D 14:22:06.069147] (0x5e1b70) getconfig state: init
>>[D 14:22:06.069162] (0x5e1b70) getconfig (FR sm) state: release:
>>(error_code = 0)
>>[D 14:22:06.069179] (0x5e1b70) getconfig (FR sm) state: send_resp
>>(status = 0)
>>[D 14:22:06.069204] BMI_post_send_list: addr: 65, count: 1, total_size: 1632
>>[D 14:22:06.069216]    element 0: offset: 0x61b6f0, size: 1632
>>[D 14:22:06.069258] BMI_tcp_post_send_generic: Sent: 1632 bytes of data.
>>[D 14:22:06.069273] (0x5e1b70) getconfig (FR sm) state: cleanup
>>[D 14:22:06.069305] (0x5e1b70) getconfig state: cleanup
>>
>>[EMAIL PROTECTED] ~]# cat /tmp/pvfs2-server.log
>>[D 13:44:33.574416] PVFS2 Server version 1.4.1pre1-2006-05-25-230553
>>starting.
>>[E 13:44:37.699990]
>>PVFS2 server got signal 15 (server_status_flag: 262143)
>>[D 13:44:39.722231] PVFS2 Server version 1.4.1pre1-2006-05-25-230553
>>starting.
>>[E 13:47:29.145477]
>>PVFS2 server got signal 15 (server_status_flag: 262143)
>>[D 13:47:31.168877] PVFS2 Server version 1.4.1pre1-2006-05-25-230553
>>starting.
>>[D 13:47:31.169626] Passing tcp://node2:3334 as BMI listen address.
>>[D 13:47:31.169682] BMI_tcp_initialize: Initializing TCP/IP module.
>>[D 13:47:31.169755] BMI_tcp_initialize: TCP/IP module successfully
>>initialized.
>>[D 13:47:31.171265] dbpf_thread_initialize: initialized
>>[D 13:47:31.172522] collection lookup: version is 0.1.2
>>[D 13:47:31.172670] - set handle re-use timeout to 360 seconds (ret=0)
>>[D 13:47:31.172826] dbpf_thread_function started
>>[D 13:47:31.172883] File system pvfs2-fs using handles:
>>        1171354720-1561806291
>>[D 13:47:31.172895] Sync on metadata update for pvfs2-fs: yes
>>[D 13:47:31.172909] Sync on I/O data update for pvfs2-fs: no
>>[D 13:47:31.172942] Storage Init Complete (aio-threaded)
>>[D 13:47:31.172953] 1 filesystem(s) initialized
>>[D 13:47:31.173461] Initialization completed successfully.
>>
>>
>>
>>[EMAIL PROTECTED] pvfs2]# cat /etc/pvfs2-fs.conf
>><Defaults>
>>        UnexpectedRequests 50
>>        LogFile /tmp/pvfs2-server.log
>>        EventLogging storage,network,server
>>        LogStamp usec
>>        BMIModules bmi_tcp
>>        FlowModules flowproto_multiqueue
>>        PerfUpdateInterval 1000
>>        ServerJobBMITimeoutSecs 30
>>        ServerJobFlowTimeoutSecs 30
>>        ClientJobBMITimeoutSecs 300
>>        ClientJobFlowTimeoutSecs 300
>>        ClientRetryLimit 5
>>        ClientRetryDelayMilliSecs 2000
>></Defaults>
>>
>><Aliases>
>>        Alias node1 tcp://node10:3334
>>        Alias node10 tcp://node11:3334
>>        Alias node11 tcp://node1:3334
>>        Alias node2 tcp://node2:3334
>>        Alias node3 tcp://node3:3334
>>        Alias node4 tcp://node4:3334
>>        Alias node5 tcp://node5:3334
>>        Alias node6 tcp://node6:3334
>>        Alias node7 tcp://node7:3334
>>        Alias node8 tcp://node8:3334
>>        Alias node9 tcp://node9:3334
>></Aliases>
>>
>><Filesystem>
>>        Name pvfs2-fs
>>        ID 833677876
>>        RootHandle 1048576
>>        <MetaHandleRanges>
>>                Range node1 4-390451575
>>        </MetaHandleRanges>
>>        <DataHandleRanges>
>>                Range node10 390451576-780903147
>>                Range node11 780903148-1171354719
>>                Range node2 1171354720-1561806291
>>                Range node3 1561806292-1952257863
>>                Range node4 1952257864-2342709435
>>                Range node5 2342709436-2733161007
>>                Range node6 2733161008-3123612579
>>                Range node7 3123612580-3514064151
>>                Range node8 3514064152-3904515723
>>                Range node9 3904515724-4294967295
>>        </DataHandleRanges>
>>        <StorageHints>
>>                TroveSyncMeta yes
>>                TroveSyncData no
>>                AttrCacheKeywords datafile_handles,metafile_dist
>>                AttrCacheKeywords dir_ent, symlink_target
>>                AttrCacheSize 4093
>>                AttrCacheMaxNumElems 32768
>>        </StorageHints>
>></Filesystem>
>>
>>
>>    
>>
>
>
>  
>

_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users

Reply via email to