Red Hat Bugzilla – Bug 187826
File system corruption on multipath device
Last modified: 2010-01-11 21:24:54 EST
Description of problem: multipathed filesystem become corrupt when heavy I/O occurs. Version-Release number of selected component (if applicable): kernel: 2.6.9-11.ELsmp #1 SMP Fri May 20 18:25:30 EDT 2005 x86_64 x86_64 x86_64 GNU/Linux Red Hat Enterprise Linux AS release 4 (Nahant Update 1) Fibre Channel: QLogic Corp. QLA2300 64-bit Fibre Channel Adapter device-mapper-multipath-0.4.5-6.0.RHEL4 How reproducible: Very Steps to Reproduce: 1. create fs using mkfs.ext3 and mount /dev/mapper/lun10p1 onto /local 2. touch file on /local-writes are fine 3. run script to generate multiple folders/files in /local Actual results: FS is corrupt and mounted as read-only Expected results: No corruption Additional info: /var/log/ messages Apr 1 04:00:02 proddb2 kernel: device-mapper: dm-multipath: Failing path 8:64. Apr 1 04:00:02 proddb2 kernel: device-mapper: dm-multipath: Failing path 65:112. Apr 1 04:00:02 proddb2 kernel: EXT3-fs error (device dm-16): ext3_get_inode_loc: unable to read inode block - inode=16695297, block=33390594 Apr 1 04:00:02 proddb2 kernel: Aborting journal on device dm-16. Apr 1 04:00:02 proddb2 kernel: Buffer I/O error on device dm-16, logical block 1545
Sorry for the late response. Is this still a problem? If I had to guess based on the snippit of /var/log/messages you provided, the load caused an all-paths-down situation and queue_if_no_path was not set, so the I/O error was returned to the filesystem and you got corruption. Or there could have been another bug somewhere.
It was an issue with the specific driver on the QLogic card. No longer an issue.