Hi, I have a few questions about OCFS2 administration and features : - I have the following architecture (simple, first step : a one node cluster...): - a iSCSI server - a client, using the iSCSI exported device with an OCFS2 filesystem ocfs2 is version 1.1.2, linux kernel is version 2.6.12.3 . When I run a benchmark (using iozone) on the client on the mounted ocfs2 fs the kernel panic when reaching large file. => "ocfs2 is very sorry to be fencing this system by panicing". In order to end benchmarking i changed the O2CB_HEARTBEAT_THRESHOLD from 7 to O2CB_HEARTBEAT_THRESHOLD=3001, this way the benchmark can finish without problem. Is there a way/a feature to avoid panicing, and instead get I/O errors on system calls using the OCFS2 fs ? How to get information about all nodes of the cluster (status of each node, quorum, DLM usage)? thanks.
Sunil Mushran
2005-Oct-18 13:20 UTC
[Ocfs2-users] OCFS2 - panic behaviour and cluster status
1.1.2 is fairly old in the 1.1.x stream. Use 1.1.6. You are hitting a known bug. spgle wrote:> Hi, > I have a few questions about OCFS2 administration and features : > > - I have the following architecture (simple, first step : a one node > cluster...): > - a iSCSI server > - a client, using the iSCSI exported device with an OCFS2 filesystem > > ocfs2 is version 1.1.2, linux kernel is version 2.6.12.3 . > > When I run a benchmark (using iozone) on the client on the mounted > ocfs2 fs the kernel > panic when reaching large file. > => "ocfs2 is very sorry to be fencing this system by panicing". > > In order to end benchmarking i changed the O2CB_HEARTBEAT_THRESHOLD > from 7 > to O2CB_HEARTBEAT_THRESHOLD=3001, this way the benchmark can finish > without problem. > > Is there a way/a feature to avoid panicing, and instead get I/O errors > on system calls > using the OCFS2 fs ? > > How to get information about all nodes of the cluster (status of each > node, > quorum, DLM usage)? > > thanks. > > > > _______________________________________________ > Ocfs2-users mailing list > Ocfs2-users@oss.oracle.com > http://oss.oracle.com/mailman/listinfo/ocfs2-users
I upgraded to release 1.1.6, using ocfs2-tools revision 1100 from subversion repository. Still have the same problem, mounted FS : /dev/sda1 on /mnt/OCFS2 type ocfs2 (rw,_netdev,heartbeat=local) Context : ocfs2 : 1.1.6 ocfs2-tools : last subversion revision 1100 Line screened on panic : (6,0): o2hb_write_timeout:164 ERROR: Hearbeat write timeout to device sda1 after 12000 milliseconds (6,0): o2hb_stop_all_regions:1727 ERROR: stooping hearbeat on all active regions. Kernel panic - not syncing : ocfs2 is verry sorry to be fencing this system by panicing. /dev/sda1 is an iSCSI device. Sunil Mushran wrote:> 1.1.2 is fairly old in the 1.1.x stream. Use 1.1.6. > You are hitting a known bug. > > spgle wrote: > >> Hi, >> I have a few questions about OCFS2 administration and features : >> >> - I have the following architecture (simple, first step : a one node >> cluster...): >> - a iSCSI server >> - a client, using the iSCSI exported device with an OCFS2 filesystem >> >> ocfs2 is version 1.1.2, linux kernel is version 2.6.12.3 . >> >> When I run a benchmark (using iozone) on the client on the mounted >> ocfs2 fs the kernel >> panic when reaching large file. >> => "ocfs2 is very sorry to be fencing this system by panicing". >> >> In order to end benchmarking i changed the O2CB_HEARTBEAT_THRESHOLD >> from 7 >> to O2CB_HEARTBEAT_THRESHOLD=3001, this way the benchmark can finish >> without problem. >> >> Is there a way/a feature to avoid panicing, and instead get I/O errors >> on system calls >> using the OCFS2 fs ? >> >> How to get information about all nodes of the cluster (status of each >> node, >> quorum, DLM usage)? >> >> thanks. >> >> >> >> _______________________________________________ >> Ocfs2-users mailing list >> Ocfs2-users@oss.oracle.com >> http://oss.oracle.com/mailman/listinfo/ocfs2-users > > > >