Hello list,

I've got a problem with a 2-node active/passive setup, running with heartbeat 
2.0.7 on Novell SuSE 10.1.
The cluster is running in R1-style configuration, with a drbd-disk, IPaddress, 
IPsrcaddr, nmb, smb, nfs and one proprietary
application as highly available services.
As heartbeat-media the nodes use both eth0 (via switch) and eth1 (direct-link) 
network-interfaces (unicast), and a direct serial
connection. The netspeed is 1Gbit/s full duplex for both interfaces, the serial 
line works at 57600 baud.
Furthermore, I've declared three IP-Adresses as ping nodes.

Starting the cluster and running the resources is all fine, manual switching 
between nodes with hb_takeover and hb_standby
works as expected, with all resources being started respectively stopped as 
they should.

But the cluster shows some rather weird runtime behaviour. On a regular basis, 
both nodes report their eth0-network interfaces
being down, therefore reporting their ping-group as dead. As this doesn't 
happen exactly synchronized, it sometimes provokes a
resource failover, depending on which node declared itself dead in the first 
place.
This state lasts for about three seconds, after that time both nodes fire their 
eth0-interfaces back up and resume working as if
nothing happened. The network-interfaces use hardware from Marvell and are 
driven by the sk98lin-driver.

I hope anyone has got an idea about this.

Regards,
Ronald

Mit freundlichen Grüßen

BBR Verkehrstechnik GmbH
i.A. Ronald Unterreiner
- Projektabwicklung Informationssysteme -

--
BBR - Baudis Bergmann Rösch - Verkehrstechnik GmbH
Pillaustraße 1e
D - 38126 Braunschweig

T: +49.531.27300.675
F: +49.531.27300.940
M: +49.170.5777527
@: [EMAIL PROTECTED]
W: www.bbr-vt.de

Registergericht: AG Braunschweig, HRB 3037

Geschäftsführer:
Dipl.-Ing. Arne Baudis
Dipl.-Ing. Thomas Bergmann
Dipl.-Ing. Frank-Michael Rösch

USt.-ID-Nr.: DE 114 877 881

_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to