>>>>> "Dave" == Dave Kleikamp <[EMAIL PROTECTED]> writes:
> On Tuesday 09 July 2002 05:23, Bernhard Mogens Ege wrote: >> We have just replaced our Raid controller with a new one, expecting >> the errors to subside, but they keep coming: >> >> Jul 9 02:48:57 boston kernel: find_entry called with index = 0 >> Jul 9 04:02:22 boston kernel: find_entry called with index = 0 >> >> We did a full fsck after the installation of the new raid controller, >> and we now believe our problems are JFS related. I guess JFS just >> isn't NFS friendly enough yet (as written in the FAQ. Should we >> switch filesystem or could we perform additional tests? > Unfortunately, fsck.jfs does not currently fix the problem leading to > the find_entry errors. However, these warnings should only lead to > problems reading and accessing entries in the directory, and should not > cause further problems. Ok. >> The above log entries are usually followed up by numerous other JFS >> errors after a while and then the server crashes (after at most 5-7 >> days with the newest JFS/kernel release). > Are these errors similar to the ones reported in your last message? Yes. They do vary a bit. When the system has crashed, I may be able to log in rmeotely, sometimes on the console only and sometimes not at all. If I can log in, acessing the JFS partition causes the shell to hang (as does the sync command). The ext2 system filesystem remains in working condition. > Those error messages indicated massive filesystem corruption, as if > entire portions of the partition where corrupted. I'm not aware of any > other reported problems that indicate this kind of problem. Also got this message today (it usually turns up): Jul 9 12:21:05 boston kernel: diRead: di_nlink is zero. ino=24742 I have recently made a parity check on the raid5 but the parity seems in order. > I can try to stress JFS under NFS. I haven't played with NFS in a > while. I assume you're using NFSv3? What kind of workload is being > run on the clients? We are exporting the JFS filesyste via NFSv3 and Samba. The server also handles backup of the entire local (ext2,jfs) filesystem (afbackup). We have about 10 clients that acces either using NFSv3 or Samba (windows). Large files (over 500Mb) are often handled as is small files (compile projects and regular homedir usage). I don't think we have large than 1GB files. Bernhard Ege _______________________________________________ Jfs-discussion mailing list [EMAIL PROTECTED] http://www-124.ibm.com/developerworks/oss/mailman/listinfo/jfs-discussion
