Cannot configure LLT if one of the LLT link is faulty.

book

Article ID: 100020473

calendar_today

Updated On:

Description

Error Message

# /sbin/lltconfig -c
LLT lltconfig ERROR V-14-2-15336 dlatt failed: DLerror 4 - Invalid argument
LLT lltconfig ERROR V-14-2-15074 cannot DL_ATTACH 8 on /dev/bge 
 

Cause

Prior to VCS  5.1 release:

This behavior is as per LLT design. Currently even if one of the LLT links is faulty, LLT does not get configured.
 

From VCS  5.1 release:

On failure of configuration of one link, lltconfig continues to configure the remaining links and if it succeeds to configure even one link, the output of "lltconfig" command is shown as "LLT is running" and LLT works fine. The output of "lltstat -l" command shows the info of the link which succeeded in configuring.

 

Resolution

If one of the LLT heartbeat links is faulty (either due to hardware failure or due to misconfiguration), LLT does not get configure and would not start.
 
Consider below /etc/llttab as an example:
 
set-node node1
set-cluster 7
link bge1 /dev/bge:1 - ether 0xCAFE -
link-lowpri bge0 /dev/bge:0 - ether - -
link bge8 /dev/bge:8 - ether 0xCAFF -           <=== This is faulty (non-existent interface)
 
When the node is rebooted or '/sbin/lltconfig -c' is manually run, it errors out with:
 
# /sbin/lltconfig -c
LLT lltconfig ERROR V-14-2-15336 dlatt failed: DLerror 4 - Invalid argument
LLT lltconfig ERROR V-14-2-15074 cannot DL_ATTACH 8 on /dev/bge 

 
 
# lltconfig
LLT is not running
LLT can not communicate with other node.
 
Other ways to check individual interface:
- On failed interface:
===================
# ./getmac /dev/qfe:6
LLT getmac ERROR V-14-2-15336 dlatt failed: DLerror 8 - Error 0
getmac: cannot DL_ATTACH 6 to /dev/qfe
===================
 
 - On working interface:
===================
# ./getmac /dev/qfe:2
/dev/qfe:2      08:00:20:E2:C7:42
===================
This behavior is as per LLT design. Currently even if one of the LLT links is faulty, LLT does not get configured.
 
Either fix the hardware issue on the defective network interface or modify /etc/llttab and comment out the problematic interface.
Once this is fixed, either reboot the node or manually start LLT, GAB and VCS.
 
This behavior is enhanced in the Veritas Cluster Server (VCS) 5.1 version. 
From VCS  5.1, on failure of configuration of one link, lltconfig continues to configure the remaining links and if it succeeds to configure even one link, the output of "lltconfig" command is shown as "LLT is running" and LLT works fine. The output of "lltstat -l" command shows the info of the link which succeeded in configuring.
 

Sample output from VCS 5.1 system is shown below:

# lltconfig
LLT is not running
# cat /etc/llttab
set-node node1
set-cluster 7
link e1000g0 /dev/e1000g:0 - ether - -
link e1000g2 /dev/e1000g:2 - ether - -
# /lib/svc/method/llt start
Starting LLT...
LLT lltconfig ERROR V-14-2-15336 dlatt failed: DLerror 4 - Invalid argument
LLT lltconfig ERROR V-14-2-15074 cannot DL_ATTACH 2 on /dev/e1000g
Starting LLT done.
# lltconfig
LLT is running
# lltstat -l
LLT link information:
link 0  e1000g0 on etherfp hipri
        mtu 1500, sap 0xcafe, broadcast FF:FF:FF:FF:FF:FF, addrlen 6
        txpkts 0  txbytes 0
        rxpkts 284  rxbytes 17024
        latehb 0  badcksum 0  errors 0
#

Applies To

This is applicable to VCS across all platforms.

Issue/Introduction

Cannot configure LLT if one of the LLT link is faulty.

Additional Information

ETrack: 1476252