Sun Enterprise 10000 Server
 
Sun Fire 12K Server
 
Sun Fire 15K Server
 

PLA-Support>Sun Systems>SPARC>Enterprise>SN-SPARC: SF-Exxk
 
.Old GCS Categories>Sun Microsystems>Servers>High-End Servers
 

Applies to:

Purpose

Troubleshooting Initial Local Network Issues on the Sun Enterprise[TM]10000/SSP or Sun Fire[TM]15K/System Controller.

Troubleshooting Steps

Description
SSP functions may fail to work under the following circumstances :

  - The FIRST time a new, never-before running SSP is brought online
  - The FIRST  attempt to switchover to a spare, untested SSP
  - The FIRST time a SSP is rebooted right after the customer changes IP addresses and/or netmasks associated with the SSP and domains
  - Directly after the customer restores a ssp_backup archive from an unverified SSP

This may be an indication of a configuration/network-related problem.

Steps to Follow
Generic Symptoms :
   1. SSP failover appears to work or partially work
   2. CB reset may not complete
   3. Unable to do netcon, power, bringup, other commands
   4. Possible error messages related to snmp in $SSPVAR/adm/messages
   5. Problems generally are NOT seen on SF12K/15K platform when using default netmasks for SC and internal network interfaces

Troubleshooting Hints :

  1. Identify which subnet masks and IP addresses are used for the     network interfaces experiencing  the problems. Verify  that VALID subnet masks, and IP addresses within that subnet mask, are being used. Compare the information collected with the tables  of valid subnet mask/ IP addresses  found in this document.

  2. There are some host addresses reserved for special uses.  In all network classes, host numbers  0 and 255 are reserved. An IP address with all host   bits set to 0 identifies the network itself.  An IP  address with all the host bits set to 1 is a broadcast address.  The point here is to NOT use a host address with bits set to all 0's or all  1's, especially when submasks are being used.

  3. Not all subnet masks are straight forward to work with in determining what is a permissible network number and host number. Refer to RFC 1878 for a list of all possible subnet masks and the valid
address they produce.

     (see http://ietf.org/rfc.html to obtain specific RFCs)

  4. In order to function properly, every interface on a specific physical network segment must have the same subnet mask.  Check the values in /etc/netmasks on both the working and problematic SSP
to insure these netmasks match.

  5. Check output of ifconfig -a to verify broadcast address used (this value is calculated by ifconfig by default - and may cause network issues if invalid IP addresses for a given subnet mask are initially used)

  6. Two common interface configuration problems are misconfigured subnet masks and incorrect IP addresses that are INVALID within that subnet mask.

  7. A bad subnet mask is indicated when the host can reach other hosts on it's local subnet and remote hosts on distant networks, but it CANNOT reach hosts on other LOCAL subnets.

  8. An incorrectly set IP address can be a subtle problem. If the network part of the address is incorrect, every ping will fail with a "no response" error. If the host part of the address is wrong, the problems can be more difficult to detect. Additionally, your system may NOT be the one that is misconfigured.  These types of problems cannot be readily discovered using ifconfig.

  9. The following tables are excerpted from RFC 1878 to give a quick reference as to valid netmask/hostID numbering conventions. Verify your customer is using a VALID IP address for the subnet mask used :

     (Excerpt from http://www.ietf.org/rfc/rfc1878.txt)

     RFC 1878             Subnet Table               December 1995


Subnets and Networks

   The number of available network and host addresses are derived from
   the number of bits used for subnet masking.  The tables below depict
   the number of subnetting bits and the resulting network, broadcast
   address, and host addresses.  Please note that all-zeros and all-ones
   subnets are included in Tables 1-1 and 1-2 per the current,
   standards- based practice for using all definable subnets [4].

   Table 1-1 represents traditional subnetting of a Class B network
   address.


Subnet Mask     # of nets    Net. Addr.  Host Addr Range  Broadcast Addr.
Bits of Subnet  hosts/subnet

255.255.254.0   128 nets      N.N.0.0     N.N.0-1.N        N.N.1.255
7 bit subnet    510           N.N.2.0     N.N.2-3.N        N.N.3.255
                              N.N.4.0     N.N.4-5.N        N.N.5.255
                              N.N.250.0   N.N.250-251.N    N.N.251.255
                              N.N.252.0   N.N.252-253.N    N.N.253.255
                              N.N.254.0   N.N.254.N        N.N.254.255

Subnet Mask     # of nets    Net. Addr.  Host Addr Range  Broadcast Addr.
Bits of Subnet  hosts/subnet

255.255.255.0   255 nets      N.N.0.0     N.N.0.N          N.N.0.255
8 bit subnet    253           N.N.1.0     N.N.1.N          N.N.1.255
                              N.N.252.0   N.N.252.N        N.N.252.255
                              N.N.253.0   N.N.253.N        N.N.253.255


References:
These documents will provide further troubleshooting steps:

• Sun Enterprise[TM] 10000: Troubleshooting Control Board Network Issues (Doc ID 1321300.1)
• Sun Fire[TM] 12K/15K/E20K/E25K: Management Networks (MAN) (Doc ID 1018896.1)

Internal Section

Internal only References:

• Sun Fire[TM] 12K/15K: Troubleshooting the I1 MAN Network (Doc ID 1010314.1)

Keywords: E10K, E10000, SSP, SF15K, SF12K, NETMASK, SUBNET
Previously Published As 72292

References

This solution has no attachment