[DRBD-user] Re: filesystem corruptions

Helmut Wollmersdorfer helmut.wollmersdorfer at gmx.at
Mon Oct 10 00:25:11 CEST 2005

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.

Lars Ellenberg wrote:

> an other nice test suggested three years ago by people encountering the
> infamous via dma northbridge bug
> (or was it south? I always confuse them...)

The southbridge does DMA and PCI stuff.

> # cat /dev/hde > null & # or some scsi device
> # cat /dev/hdg > null & # or wherever your disks are
> # ping -f -s 64000 otherMachine &

With kernel 2.4.18 I needed only ssh and scp to freeze my VIA EPIA 
(southbridge VT8235). Since the workarounds of Alan Cox around kernel 
2.4.21/22/23 they are gone - mostly.

> one of the workaround "solutions" back then was to place the NIC
> into some other pci slot :-)
> ... hardware ... sometimes I wonder why it works most of the time.

Read the longest thread on viarena about workarounds
EPIA-M: DMA & VT8235 Lock-up

A very curious is to run memtest in the background with low priority to 
avoid low temperatures.

Hmm ..., a little bit more on topic and DRBD related, my standby node 
freezed from time to time, especially short after leaving to the weekend 
and turning down the room temperature ...

Both nodes have exactly the same hardware, same firmware, same software. 
Node1 runs without problems since 2.5 years. The only difference is, 
that Node1 is the primary and has a little bit more to do.

In August I changed mainboard, memory and case (power supply). Again, 
Node2 stopped shortly after leaving to my holydays in Italy (after ~3 
weeks uptime). Now it runs since three weeks.

Hardware ...

Helmut Wollmersdorfer

More information about the drbd-user mailing list