Michael Oberg
2006-Nov-20 17:26 UTC
[Lustre-discuss] PPC64 client to x86_64 cluster, version 1.5.95 produces LustreError on MDT
Hello, Our Lustre filesystem mounts fine on a x86_64 client, but I am encountering the following error when attempting to mount from a PPC64 client (mds01 is an alias for our MDS): # mount -t lustre mds01@tcp0:/testfs /mnt/testfs mount.lustre: mount 172.30.2.11@tcp:/testfs at /mnt/testfs failed: Input/output error Is the MGS running? On the MDS, /var/log/messages shows the following: Nov 20 17:02:38 ma0121en LustreError: 7180:0:(pack_generic.c:710:lustre_unpack_msg_v2()) message length 168 too small for 50331648 buflens Nov 20 17:02:38 ma0121en LustreError: 7180:0:(service.c:553:ptlrpc_server_handle_request()) error unpacking request: ptl 26 from 12345-172.30.1.104@tcp xid 43 Nov 20 17:03:03 ma0121en LustreError: 7183:0:(pack_generic.c:710:lustre_unpack_msg_v2()) message length 96 too small for 16777216 buflens Nov 20 17:03:03 ma0121en LustreError: 7183:0:(service.c:553:ptlrpc_server_handle_request()) error unpacking request: ptl 26 from 12345-172.30.1.104@tcp xid 44 Our configuration is as follows: Storage Cluster: - 12 OSS''s, with 24 OST''s, 118TB total, with 2.4TB for the MDT - Dual Opteron 248''s, 4GB RAM - 2-way bonded gige - SuSE SLES9 running kernel 2.6.5-7.282-smp Client: - Openpower 720 with 4 SMT enabled Power5 processors - SuSE SLES9 running 2.6.5-7.282-pseries64 kernel, with patch attached to bug 10634 - 8-way bonded gige Both were built using the same 1.5.95 source: Cluster: ./configure --prefix-/opt/lustre-1.5.95 --with-linux=/usr/src/linux-2.6.5-7.282 --with-linux-obj=/usr/src/linux-2.6.5-7.282-obj/x86_64/smp/ Client: ./configure --with-linux=/usr/src/linux-2.6.5-7.282-lustre --disable-liblustre --disable-doc --disable-openibnal --disable-readline --disable-test Any input or debugging suggestions would be appreciated, Michael Oberg Research Systems Evaluation Team National Center for Atmospheric Research (NCAR) Office: 303.497.1268, Cell: 720.938.6585 oberg@ucar.edu
Nathaniel Rutman
2006-Nov-20 17:46 UTC
[Lustre-discuss] PPC64 client to x86_64 cluster, version 1.5.95 produces LustreError on MDT
There is a problem with mixed-endian installations (bug 11214) https://bugzilla.clusterfs.com/show_bug.cgi?id=11214 We''re working on it, sorry. Michael Oberg wrote:> Hello, > > Our Lustre filesystem mounts fine on a x86_64 client, but I am > encountering the following error when attempting to mount from a PPC64 > client (mds01 is an alias for our MDS): > > # mount -t lustre mds01@tcp0:/testfs /mnt/testfs > mount.lustre: mount 172.30.2.11@tcp:/testfs at /mnt/testfs failed: > Input/output error > Is the MGS running? > > On the MDS, /var/log/messages shows the following: > > Nov 20 17:02:38 ma0121en LustreError: > 7180:0:(pack_generic.c:710:lustre_unpack_msg_v2()) message length 168 > too small for 50331648 buflens > Nov 20 17:02:38 ma0121en LustreError: > 7180:0:(service.c:553:ptlrpc_server_handle_request()) error unpacking > request: ptl 26 from 12345-172.30.1.104@tcp xid 43 > Nov 20 17:03:03 ma0121en LustreError: > 7183:0:(pack_generic.c:710:lustre_unpack_msg_v2()) message length 96 too > small for 16777216 buflens > Nov 20 17:03:03 ma0121en LustreError: > 7183:0:(service.c:553:ptlrpc_server_handle_request()) error unpacking > request: ptl 26 from 12345-172.30.1.104@tcp xid 44imeouts were required > > Our configuration is as follows: > > Storage Cluster: > - 12 OSS''s, with 24 OST''s, 118TB total, with 2.4TB for the MDT > - Dual Opteron 248''s, 4GB RAM > - 2-way bonded gige > - SuSE SLES9 running kernel 2.6.5-7.282-smp > > Client: > - Openpower 720 with 4 SMT enabled Power5 processors > - SuSE SLES9 running 2.6.5-7.282-pseries64 kernel, with patch attached > to bug 10634 > - 8-way bonded gige > > Both were built using the same 1.5.95 source: > > Cluster: ./configure --prefix-/opt/lustre-1.5.95 > --with-linux=/usr/src/linux-2.6.5-7.282 > --with-linux-obj=/usr/src/linux-2.6.5-7.282-obj/x86_64/smp/ > > Client: ./configure --with-linux=/usr/src/linux-2.6.5-7.282-lustre > --disable-liblustre --disable-doc --disable-openibnal --disable-readline > --disable-test > > Any input or debugging suggestions would be appreciated, > > Michael Oberg > Research Systems Evaluation Team > National Center for Atmospheric Research (NCAR) > Office: 303.497.1268, Cell: 720.938.6585 > oberg@ucar.edu > > > > > > > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > >