HP Integrity rx4640 - Application log complaining OS Core get host by addr failed

[javasoup]

Apps & end-usrs are complaining that their dialog connections will hang sometimes.
When I check the applications log, there are occurrences of
Operating system call gethostbyaddr failed.

========================================
Summary
0/2/1/1 - connected to the MSL6000
fc 1 0/2/1/1 fcd CLAIMED INTERFACE HP A6826-60001 2Gb Dual Port PCI/PCI-X Fibre Channel Adapter (FC Port 2)
/dev/fcd1

fc 3 0/4/2/1 fcd CLAIMED INTERFACE HP A6826-60001 2Gb Dual Port PCI/PCI-X Fibre Channel Adapter (FC Port 2)
/dev/fcd3

==============================================

All FC Cards seemed alright. Even with commands like these:

# fcmsutil /dev/fcd1
# fcmsutil /dev/fcd1 stat
# fcmsutil /dev/fcd3
# fcmsutil /dev/fcd3 stat

The application is SAP based.
Could any expert kindly advise me on this problem?

Thank you very much,
Sincerely,
Edmund

[RAC_1]

Do you see any error at the same time at OS level?? Did you check syslog/dmesg?? Is nslookup back and forth happening correctly??

[Sameer_Nirmal]

Hi,

The "gethostbyaddr" call is a part of system library used for net services/protocols like tcp/ip, DNS, NFS, NIS etc.

Hence it seems that the problem lies somewhere at net services/network level. Check for any message in syslog,demsg, respective net services logs being used.
Use nslookup , netstat , nettl/netfmt etc. to make sure all OK from server and OS side.
You can involve someone from networking thereafter to check at outside network level.

[javasoup]

Yea i agree on ur pt abt checking the Network Connectivity between the server and the HP Procurve Switch.

I got the user to follow these instructions:
1. Please provide the model of the switch, is it a HP switch?
If yes, please do the following,
"supportshow" output from any switches connected to the VA

2 ways for customer to collect supportshow 1. Direct connection using Notebook to the switch,
key in the login and passwd, default "admin" for both.
"supportshow" is the command, and do a screen capture / logging

2. Do a telnet to the switch, "supportshow" and do a screen capture / logging.



BUT, he told me "supportshow" is an invalid command on his side which really got me thinking again what is wrong.

Any idea?

Thanks!
Sincerely,
Edmund

[RAC_1]

On server side, if you experienced any network errors or not.
Check syslog.log
dmesg
netfmt -f /var/adm/nettl.LOGxx
lanadmin -g "nm_id_of_card"
nslookup on ip and on name
ping

[rick jones]

Gethostbyaddr() is used to map an IP address to a hostname. As already pointed-out it has nothing to do with fibre channel mass storage :)

Does the error message say anything about _why_ the gethostbyaddr() call failed? Or say something useful like which IP address was to be queried?

Indeed, there might be connectivity issues (in which case the caller would have spent quite a bit of time in the gethostbyaddr() call, but I'd guess that 99 times out of ten a failed gethostbyaddr() call stems from there not being a so called "PTR" record in the DNS to map the IP address to the hostname.

That can be for any number of reasons:

1) the hostmaster forgot to edit the zone file correctly when adding a new host

2) someone is using a "guessed" rather than assigned IP address

If you happen to run tcpdump on the host, looking for traffic to/from port 53, you may see the failed PTR queries and so know for which IP address(es) they are failing.

If the log includes the rather useful bit of info such as the IP address, you could try an nslookup to see if there is indeed a PTR record in the DNS. (or use the dig utility instead of nslookup)

Drifting a bit, with the advent of IPv6 and such, the gethostbyaddr()/gethostbyname() calls are becoming obsolete. Software should be migrating to getnameinfo() and getaddrinfo() instead as this will be one of the necessary steps towards supporting IPv6