Bug #4204
closedkclient: regression triggered by direct io path
0%
Description
this probably related to recent changes in the testing branch (which fails); master passes.
2013-02-19T01:20:29.912 INFO:teuthology.task.workunit.client.0.out:writing pattern 2013-02-19T01:20:29.912 INFO:teuthology.task.workunit.client.0.out:read_direct buf_align 0 offset 4190208 len 1024 2013-02-19T01:20:29.912 INFO:teuthology.task.workunit.client.0.out:read_sync buf_align 0 offset 4190208 len 1024 2013-02-19T01:20:29.913 INFO:teuthology.task.workunit.client.0.out:read_direct buf_align 0 offset 4190208 len 2048 2013-02-19T01:20:29.913 INFO:teuthology.task.workunit.client.0.out:read_sync buf_align 0 offset 4190208 len 2048 2013-02-19T01:20:29.913 INFO:teuthology.task.workunit.client.0.out:read_direct buf_align 0 offset 4190208 len 4096 2013-02-19T01:20:29.913 INFO:teuthology.task.workunit.client.0.out:read_sync buf_align 0 offset 4190208 len 4096 2013-02-19T01:20:29.913 INFO:teuthology.task.workunit.client.0.out:read_direct buf_align 0 offset 4190208 len 8192 2013-02-19T01:20:29.914 INFO:teuthology.task.workunit.client.0.out:read_sync buf_align 0 offset 4190208 len 8192 2013-02-19T01:20:29.914 INFO:teuthology.task.workunit.client.0.out:read_direct buf_align 0 offset 4190208 len 16384 2013-02-19T01:20:29.914 INFO:teuthology.task.workunit.client.0.out:read_sync buf_align 0 offset 4190208 len 16384 2013-02-19T01:20:29.914 INFO:teuthology.task.workunit.client.0.out:read_direct buf_align 0 offset 4190720 len 1024 2013-02-19T01:20:29.915 INFO:teuthology.task.workunit.client.0.out:read_sync buf_align 0 offset 4190720 len 1024 2013-02-19T01:20:29.915 INFO:teuthology.task.workunit.client.0.out:read_direct buf_align 0 offset 4190720 len 2048 2013-02-19T01:20:29.915 INFO:teuthology.task.workunit.client.0.out:read_sync buf_align 0 offset 4190720 len 2048 2013-02-19T01:20:29.915 INFO:teuthology.task.workunit.client.0.out:read_direct buf_align 0 offset 4190720 len 4096 2013-02-19T01:20:29.915 INFO:teuthology.task.workunit.client.0.out:error: offset 4190720 had 4194304
the test is:
kernel: kdb: true branch: testing nuke-on-error: true overrides: ceph: conf: osd: osd op thread timeout: 60 fs: btrfs log-whitelist: - slow request roles: - - mon.a - mon.c - osd.0 - osd.1 - osd.2 - - mon.b - mds.a - osd.3 - osd.4 - osd.5 - - client.0 tasks: - chef: null - clock: null - ceph: null - kclient: null - workunit: clients: all: - direct_io
Updated by Alex Elder about 11 years ago
I'm certain it's because of this commit: 29c3c8b721a96b4a82f4224527c7103e5e910b80
Which was done for this issue: http://tracker.ceph.com/issues/4166
I'm just pulling that commit out and am going to abandon that
change completely. It was not well thought out and I didn't
test it adequately, and I'm ashamed of myself.
I'm testing the result now and will update the testing branch
accordingly after I've verified this is fixed.
Updated by Alex Elder about 11 years ago
I'm having a lot of trouble testing this.
I can write to files in a ceph file system. But direct I/O
writes gives ENODEV.
I can read from files in a ceph file system. But direct I/O
read again produces ENODEV.
This is the case with the testing branch, but with the master
branch as well. The results below are from the master branch.
root@plana92:/tmp/cephtest# dd if=/dev/zero of=/tmp/cephtest/mnt.0/foo bs=65536 count=16
16+0 records in
16+0 records out
1048576 bytes (1.0 MB) copied, 0.0593075 s, 17.7 MB/s
root@plana92:/tmp/cephtest# dd if=/tmp/cephtest/mnt.0/foo bs=65536 of=/tmp/zero dd: reading `/tmp/cephtest/mnt.0/foo': No such device or address
16+0 records in
16+0 records out
1048576 bytes (1.0 MB) copied, 0.00306812 s, 342 MB/s
root@plana92:/tmp/cephtest# dd if=/dev/zero of=/tmp/cephtest/mnt.0/foo bs=65536 count=16 oflag=direct
dd: writing `/tmp/cephtest/mnt.0/foo': No such device or address
1+0 records in
0+0 records out
0 bytes (0 B) copied, 0.127708 s, 0.0 kB/s
root@plana92:/tmp/cephtest# dd if=/tmp/cephtest/mnt.0/foo bs=65536 of=/tmp/zero iflag=direct
dd: reading `/tmp/cephtest/mnt.0/foo': No such device or address
0+0 records in
0+0 records out
0 bytes (0 B) copied, 0.00112755 s, 0.0 kB/s
root@plana92:/tmp/cephtest#
Updated by Sage Weil about 11 years ago
- Status changed from 12 to Resolved
this is working now that alex yanked that commit.