Network question

[Patrick Chim]

Hi all,

My server is running HPUX 11.0. I ping the server from my PC and have the following results

1. When I ping using the default byte length, it is normal.
Reply from 172.19.128.11: bytes=32 time<10ms TTL=255
Reply from 172.19.128.11: bytes=32 time<10ms TTL=255
Reply from 172.19.128.11: bytes=32 time<10ms TTL=255
Reply from 172.19.128.11: bytes=32 time<10ms TTL=255
Reply from 172.19.128.11: bytes=32 time<10ms TTL=255
Reply from 172.19.128.11: bytes=32 time<10ms TTL=255


2. When I ping using larger byte length, it comes out strange words
Reply from 172.19.128.11: bytes=10000 time=3ms TTL=255
Reply from 172.19.128.11: bytes=10000 time=3ms TTL=255
Reply from 172.19.128.11: bytes=10000 time=3ms TTL=255
Reply from 172.19.128.11: bytes=10000 time=3ms TTL=255
Reply from 172.19.128.11: bytes=10000 time=3ms TTL=255
Reply from 172.19.128.11: bytes=10000 time=3ms TTL=255
Reply from 172.19.128.11: Source quench received.
Reply from 172.19.128.11: Source quench received.
Reply from 172.19.128.11: Source quench received.
Reply from 172.19.128.11: bytes=10000 time=3ms TTL=255
Reply from 172.19.128.11: bytes=10000 time=3ms TTL=255

What does 'Source quench received' means ? Is there any problem with the network card or the network buffer is low ? If so, do I need to tune the kernel parameters ?

Best regards,
Patrick

[Stefan Farrelly · Accepted solution]

These errors are totally harmless but can be a nuisance if a
ping monitor is being run on the network. There is currently no patch to remove the message but the change has been implemented in HP-UX release 11.01. It is not known at this time when 11.01 will be released. In the interim, there are two ways that these error messages can be suppressed. A single command will take care of them temporarily or the /etc/rc.config.d/nddconf file can be modified to make the change permanent.

The following command will suppress the messages until the
next time the system is reset.

# ndd -set /dev/ip ip_send_source_quench 0

A permanent solution is to add the following lines to the file
/etc/rc.config.d/nddconf:

TRANSPORT_NAME[INDEX]=ip
NDD_NAME[INDEX]=ip_send_source_quench
NDD_VALUE[INDEX]=0

Note: Substitute INDEX with a proper value depending on what other entries are in the file. See the comment in the nddconf file for advice on how to set the value of INDEX.

[Patrick Chim]

Stefan,

What is the name of the ping monitor ? As I am running MC/SG in my machine, is it the MC/SG that has some network protection ?

Regards,
Patrick

[Stefan Farrelly]

MC/SG does use some sort of ping monitor to detect if the other nodes are up/down, so does ITO or OpenView, or maybe you have some scripts someone has written doing ping tests to other devices. Whatever, it error you get is nothing to worry about. We get them sometime on our SG cluster, we ran the ndd command to prevent them.

[Carsten Krege]

Stefan, I don't know where you got your info from, but HPUX 11.01 does not and will not exist. The successor of HPUX 11.00 is HPUX 11.11 (aka HPUX 11i). I don't know whether the "problem" has been fixed in this version.

I believe by "ping monitor" you simply mean any sort of ping command that could return this message when pinging a HPUX 11.0 machine?

Patrick, SG doesn't add anything protective to the network which could confuse other applications (such as ping). The issue you mention has nothing to do with SG, but with HPUX 11.0 behaviour (also I don't know whether this is really a bug or feature??).

SG's heartbeat and network polling traffic doesn't have any influence on this.

Carsten