[DRBD-user] MD sets failing under heavy load in a DRBD/Pacemaker Cluster

Felix Frank ff at mpexnet.de
Tue Oct 4 14:57:55 CEST 2011

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


Hi,

that's an impressive cross-posting list. Somehow, I'm less than
impressed, though.

On 10/04/2011 02:01 PM, Caspar Smit wrote:
> 2) It seems to only happen when DRBD is connected (we didn't have any
> failing MD's yet when DRBD was not connected luckily!)

Have you created test scenarios in which an unconnected peer is stressed
to the same extent where you usually see crashes?

If not, I call "not DRBD related" on this one. It doesn't seem to make
any sense.

Other than that, you really want to pass up on those backport packages
and create your own, with a DRBD version that's not over a year old.

Another note: The fact that you can rebuild an array using a given disk
is no guarantee that said disk is not broken.

HTH,
Felix



More information about the drbd-user mailing list