[Ocfs2-users] Problem w/ 100Mbit heartbeat link

[Michael Steinmann]

This is while testing on two HP DL385, SLES9 x86_64 SP3 RC4
(2.6.5-7.244-smp SMP gcc-3.3), OCFS2 1.1.7-SLES.

Everything runs flawlessly when the heartbeat is running over a back2back
GigE. Copying directories with two bonnies running at the same time on
both nodes for hours didn't produce a single error.

But, we are getting errors when switching the heartbeat from the dedicated
GigE (eth2) to the service LAN (eth0, which is over a 100MBit switch with
other, unrelated traffic).

Errors while copying files (heartbeat over 100MBit):

Dec 27 13:39:17 spora01dbcb kernel:
(10415,2):ocfs2_prepare_dir_for_insert:569 ERROR: status = -28
Dec 27 13:39:17 spora01dbcb kernel: (10415,2):ocfs2_mknod:348 ERROR:
status = -28
Dec 27 13:39:17 spora01dbcb kernel:
(10415,2):ocfs2_prepare_dir_for_insert:569 ERROR: status = -28
Dec 27 13:39:17 spora01dbcb kernel: (10415,2):ocfs2_mknod:348 ERROR:
status = -28
Dec 27 13:39:17 spora01dbcb kernel:
(10415,2):ocfs2_prepare_dir_for_insert:569 ERROR: status = -28
Dec 27 13:39:17 spora01dbcb kernel: (10415,2):ocfs2_mknod:348 ERROR:
status = -28
Dec 27 13:39:17 spora01dbcb kernel:
(10415,2):ocfs2_prepare_dir_for_insert:569 ERROR: status = -28
Dec 27 13:39:17 spora01dbcb kernel: (10415,2):ocfs2_mknod:348 ERROR:
status = -28
Dec 27 13:39:17 spora01dbcb kernel:
(10415,2):ocfs2_prepare_dir_for_insert:569 ERROR: status = -28
Dec 27 13:39:17 spora01dbcb kernel: (10415,2):ocfs2_mknod:348 ERROR:
status = -28
Dec 27 13:39:17 spora01dbcb kernel:
(10415,2):ocfs2_prepare_dir_for_insert:569 ERROR: status = -28
Dec 27 13:39:17 spora01dbcb kernel: (10415,2):ocfs2_mknod:348 ERROR:
status = -28
Dec 27 13:39:17 spora01dbcb kernel:
(10415,2):ocfs2_prepare_dir_for_insert:569 ERROR: status = -28
Dec 27 13:39:17 spora01dbcb kernel: (10415,2):ocfs2_mknod:348 ERROR:
status = -28


The GigE Interface is an Intel Quad PRO/1000 MT:
0000:06:06.1 Ethernet controller: Intel Corporation 82546EB Gigabit
Ethernet Controller (rev 01)

Settings for eth2:
        Speed: 1000Mb/s
        Duplex: Full
        Port: Twisted Pair


The service LAN interface is a Broadcom BCM5704:
0000:03:06.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5704
Gigabit Ethernet (rev 10)

Settings for eth0:
        Speed: 100Mb/s
        Duplex: Full
        Port: Twisted Pair

Anyone else seeing similar issues? Ways to work around this, things to tweak?

--
mike