10-25-2013 06:26 AM
Hello All,
Recently this message started appearing on the server.
bash-2.05$ lltstat -nvv|head
LLT node information:
Node State Link Status Address
* 0 db1 OPEN
ce1 UP 00:03:BA:93:
ce6 UP 00:03:BA:85:
1 db2 OPEN
ce1 UP 00:03:BA:93:
ce6 UP 00:03:BA:95:
2 CONNWAIT
ce1 DOWN
Any advice is greatly apperciated, thank you.
10-25-2013 06:40 AM
Are either of these low-pri links?
Starting with 5.0MP3, Symantec has added the feature of one-way link detection in LLT (Etrack Incident# 1031514)
See LBN for 5.0 (Solaris) - refer to the following recommendation:
"If the following messages are seen frequently for an LLT lo-pri link, then change the LLT tunable named peertrouble to 400. Its default value is 200."
http://www.symantec.com/business/support/index?page=content&id=TECH46439
If this isn't the case, please provide the following:
- OS/platform
- VCS version
- Output of # cat /etc/llttab
- any other network related error messages seen on the cluster node(s)
10-25-2013 07:37 AM
Do you use bonded NICs?
How are the LLT links connected, crossover or through switches/vlans?
Do both links have seperat network connection?
If both use the same network this will also confuse LLT.
Are there any IP adresses for public LAN configured on the LLT links?
Please also make sure that each NIC has a unique MAC address, in your output the MAC addresses of both ce1 NICs look the same, but they are truncated so I can't tell for sure.
regards,
Dan
10-25-2013 08:44 AM
Hi,
The OS is Solaris 9 SPARC
VCS version 4.0
10-25-2013 09:12 AM
As asked before, how are the links connected? Crossover cables, intermediate hub (gotta ask), intermediate switch, ?
Is there anything else on the same fabric as ce1 that you can use to test each ce1 interface?
The messages say that node 1 link 0 has the problem. Note that, in the case of a two-node cluster, node 0 will *never* be reported as the problem, even if it really is.
First thing you can do to stop jeopardy state would be to define a link-lowpri on another interface - use UDP if needed.
Now, for troubleshooting ...
First, I would check messages file to see if you are losing physical link.
Second, test network connectivity between the two ce1 links (put IP addresses on them and do a ping test).
Third, if you've got another interface on the same ce1 fabric, try connecting to that interface ...
Fourth, now that you've got the link-lowpri in place, try new crossover cable between ce1 links.
Good luck, these are rarely easy. I have solved them more than once by buying a cheap switch at a big box store and hooking the cables up to that ... especially when the Network Group repeatedly claims "it's not us" ... until suddenly and inarguably it is ;)