Bad version message in syslog.log from cmcld

[Carlos André]

Hi all,

I have this strange message in the syslog.log file, appearing every ten seconds or so:

cmcld: Bad version 595582 cast 4

Yet the cluster seems to work fine.
We have the March 2003 HW bundle as well as the March 2003 Quality pack.
This message appears only on one node of the cluster (this is a 2-nodes cluster). Switching packages from one node to another doesn't change a thing.

Doesn anybody have an idea as to how to get rid of this message?

Thanks in advance

[Dietmar Konermann]

Hi!

I just had a quick look. The message comes from the dlpi_recv() function. It's issued when a polling packet with a wrong version id (CL_COMM_DLPI_VERSION) is received by cmcld... this should usually never happen.

Hmm, maybe there is some corruption happening due to networking problems.

What's the ServiceGuard version and patch level? I would suggest to use at least SG 11.09 and apply the latest appropriate patch... logging in this area may be more verbose/informative with recent versions.

Another good idea would be a cmapplyconf while the cluster is halted to get the configuration re-built.

Best regards...
Dietmar.

[melvyn burnard]

Hmm as Dietmat says, the only time I have seen this is when htere are networking issues.
What version of SG are you using, and what patch is installed.
Simplest way of finding this is to run:
what /usr/lbin/cmcld

[Robert-Jan Goossens]

Hi,

Are you using 10.20 ?

PHSS_17482:


1. Problem using LockManager with Restricted SAM.


2. cmcld aborts causing the node to TOC shortly after the message "Bad version # cast #" is reported in syslog.

Robert-Jan.

[Carlos André]

Ok, I've tried the cmapplyconf command, and it specifically says that there are some networking problems between the two nodes, and that it wouldn't go any further in applying the configuration because of that.
It appears I have some port aggregation problem on one node, so I am going to investigate this issue.

I'll post a new message about the outcome of the investigation to say whether it worked or not, so that people with this kind of problem can have a reference here in the ITRC.

For information, I am using HPUX B.11.00

Thank you everybody for your quick answers and have a great day!

[Carlos André]

Ok, the problem is fixed now.
It appears we had a switch misconfiguration for our nodes (ports on the switch were on auto-negociation instead of forced full-duplex). The network guy is responsable for this, he'll be punished accordingly... ;O)
Anyway, after forcing the full-duplex on the switch and rebooting the nodes, everything runs smoothly and cmapplyconf works just fine.

Thank you guys!