Broken Pipe in Solaris 11.2

[tom]

Hi,

I am seeing this issue on the dimstat agent when It tries to connect to the dimstat server

STATsrv-4135:4999> Connect -> [dimstat_Client_hostname_IPAddress] CMD= "STAT_LIST" TIME: Sun Mar  1 23:27:16 2015

STATsrv-4135:4999> Exit -> [dimstat_Client_hostname_IPAddress] (broken pipe) CMD= "STAT_LIST" TIME: Sun Mar  1 23:27:16 2015

On the dim stat server its not able to connect to agent but I can ping it, The agent status shows green in the webpage, When I start the collect above error comes

Also tried this command on the server

root@dimstatServerName:/opt/WebX/bin# ./STATcmd -h dimstat_Client_hostname -c  STAT_LIST

STAT *** NO CONNECTION

STAT *** NO CONNECTION

I also upgraded the core to  dim_STAT CoreUpdate-14 

I am currently on Solaris 11.2

 cat /etc/release 

                            Oracle Solaris 11.2 SPARC

  Copyright (c) 1983, 2014, Oracle and/or its affiliates.  All rights reserved.

                            Assembled 22 October 2014

Please help, This has never caused a issue on earlier release of solaris. 

Thanks

Tom

[Dimitri]

Hi Tom,

try to use a different IP port on the client (different from 5000
which is default) -- it's possible that some other OS services are
using it too. To change it on the client STAT-service just edit
/etc/STATsrv/STAT-service file and change 5000 port to 5001 for ex.
and restart STAT-service.. Then retry with ./STATcmd again and add
option "-p" with the new port number. Also try to execute STATcmd
locally on the machine first (from /etc/STATsrv/bin) and then remotely
from dim_STAT server..

indeed, there is something going odd..

Rgds,
-Dimitri

> --
> You received this message because you are subscribed to the Google Groups
> "dim_STAT" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to dimstat+u...@googlegroups.com.
> To post to this group, send email to dim...@googlegroups.com.
> Visit this group at http://groups.google.com/group/dimstat.
> For more options, visit https://groups.google.com/d/optout.
>

[tom]

Hi Dimitri,

I am running with 4999, I had hit 5000 earlier, I changed the port. I missed the -p option

Its now showing up the list :-)

On the agent side, The broken pipe message is still coming.

./STATcmd -h dimstat_Client_hostname -c  STAT_LIST -p 4999

STAT *** OK CONNECTION 0 sec.

STAT *** LIST COMMAND (STAT_LIST)

STAT: vmstat

STAT: mpstat

STAT: netstat

STAT: ForkExec

STAT: MEMSTAT

STAT: tailX

STAT: ioSTAT.sh

STAT: netLOAD.sh

STAT: netLOAD

STAT: psSTAT

STAT: UserLOAD

STAT: ProcLOAD

STAT: bsdlink

STAT: bsdlink.sh

STAT: sysinfo

STAT: SysINFO

STAT: Siostat

STAT: ProjLOAD

STAT: PoolLOAD

STAT: TaskLOAD

STAT: ZoneLOAD

STAT: IOpatt

STAT: CPUSet

STAT: UDPstat

STAT *** LIST END (STAT_LIST)

thanks

Tom

Ps I really love this program, Its been a  immense help in many projects.

[tom]

I have installed the dimserver on an another node and it works without any issues.

This seems to be an hardware issue particulary NIC, seeing lot NIC failures in the system messages.

Thanks for the prompt help.

Tom

[Dimitri]

Tom, everything is fine which is finishing fine ;-))

regarding "broken pipe" error messages : no need to worry too much
about them, in fact these messages were added explicitly to the
STAT-service code to help to understand for which reason the
collection of stats was stopped... And there are 2 options: the
command line output was closed unexpectedly (broken pipe) or a network
connection was closed between servers..

And both states may be normal and abnormal ;-))

When you're stopping stats collects via web interface -- all
corresponding network connections will be closed then, and
STAT-service will receive an error as "connection lost", and this will
be as expected and correct.. Now, if the client running STAT-service
will hang, or reset, or reboot -- the connection will be lost then on
the dim_STAT server side, and this will be seen as abnormal, so you'll
automatically get in your LOG messages the corresponding errors
pointing to the connection lost..

Same for "broken pipe" : generally each stat command is supposed to
loop forever, but it can be killed, or be buggy and crash (for ex.) --
in this case the STAT-service will recognize the situation and will
log a message about "broken pipe" from a corresponding stat command,
and then close the connection and allowing the dim_STAT server to
restart it in a normal way again. The only exception here is a
reserved "STAT_LIST" command -- this is a reserved order to show the
list of all available stat commands from a given STAT-service, and
this order is processed as any other stat command. And as STAT_LIST is
not supposed to loop forever, there is an "error" message about broken
pipe, once the STAT_LIST output is finished (just to say that all info
was sent to the network connection and pipe was closed)..

Well, hope it'll clarify little bit what is going inside ;-)