[Cflowd] cfdcollect stops working

Christian Cinetto cinetto@garr.it
Thu, 23 May 2002 09:28:04 +0200 (CEST)


Thanks Gonzalo for your advice, but unfortunately it does not solve my 
problem... Then I created a script which regularly checks the cflowd and 
cfdcollect status and restart all the processes if there are some 
mistakes. It seemed a good solution; so I monitored three routers and
now other errors came out!


May 21 14:15:29 nf1 cfdcollect[5671]: [I] sleeping for 275 seconds.
May 21 14:20:04 nf1 cfdcollect[5671]: [I] awakened by alarm.
May 21 14:20:05 nf1 cfdcollect[5671]: [I] connected to 127.0.0.1:2056
May 21 14:20:30 nf1 cfdcollect[5671]: [I] 127.0.0.1 has data for 3 
routers.
May 21 14:20:30 nf1 cflowd[5682]: [E] ArtsPrimitive.WriteUint16(6,1025,2) 
failed: Connection reset by peer {../include/CflowdUint16Uint16Key.hh:202}
May 21 14:20:30 nf1 cflowd[5682]: [E] PortMatrix().write(6) failed 
{CflowdCisco.cc:651}
May 21 14:20:30 nf1 cflowd[5682]: [E] g_cflowdConfig.CiscoMap().write(6) 
failed {cflowd.cc:409}
May 21 14:20:30 nf1 cflowd[5682]: [I] sent data to 127.0.0.1:35760
May 21 14:30:00 nf1 cflowd[5669]: [I] Received SIGTERM.  Exiting.
May 21 14:30:05 nf1 cflowdmux[5667]: [I] Received SIGTERM.  Exiting.
May 21 14:30:05 nf1 cfdcollect[5799]: [I] cfdcollect (version 
cflowd-2-1-b1) started with 1 cflowd instances.
May 21 14:30:05 nf1 cflowd[5797]: [I] cflowd (version cflowd-2-1-b1) 
started.
May 21 14:30:05 nf1 cflowd[5797]: [I] got semaphore: id 32769
May 21 14:30:05 nf1 cflowd[5797]: [E] 
shmget(ftok("/usr/local/arts/etc/cflowd.conf",0),20975616,SHM_R|(SHM_R>>3)|(SHM_R>>6)) 
failed: No such file or directory {CflowdPacketQueue.cc:342}
May 21 14:30:05 nf1 cflowd[5797]: [A] failed to open packet queue!  
Exiting. {cflowd.cc:991}
May 21 14:30:05 nf1 cflowdmux[5795]: [I] cflowdmux (version cflowd-2-1-b1) 
started.
May 21 14:30:05 nf1 cflowdmux[5795]: [I] created 20975616 byte packet 
queue shmem segment {CflowdPacketQueue.cc:247}
May 21 14:30:05 nf1 cflowdmux[5795]: [I] attached to 20975616 byte packet 
queue at 0x401d6000 
May 21 14:30:05 nf1 cflowdmux[5795]: [I] created semaphore: id 32769
May 21 14:30:05 nf1 cflowdmux[5795]: [I] set UDP recv queue to 261040 
bytes for fd 4 (port 8150)
May 21 14:30:05 nf1 cflowdmux[5795]: [I] set UDP recv queue to 261040 
bytes for fd 5 (port 8100)
May 21 14:30:05 nf1 cflowdmux[5795]: [I] set UDP recv queue to 261040 
bytes for fd 6 (port 8200)
May 21 14:30:06 nf1 cfdcollect[5799]: [E] connect(4,0x80e8dd4,16) (host 
127.0.0.1 port 2056) failed: Connection refused {CflowdServer.cc:112}
May 21 14:30:06 nf1 cfdcollect[5799]: [I] sleeping for 299 seconds.
May 21 14:35:04 nf1 cfdcollect[5799]: [I] awakened by alarm.
May 21 14:35:04 nf1 cfdcollect[5799]: [I] sleeping for 1 seconds.
May 21 14:35:05 nf1 cfdcollect[5799]: [I] awakened by alarm.
May 21 14:35:06 nf1 cfdcollect[5799]: [E] connect(4,0x80e8dd4,16) (host 
127.0.0.1 port 2056) failed: Connection refused {CflowdServer.cc:112}

The problem seems to be connected to cflowd 's use of the shared memory.
I tried to increase the packet buffer size to 20975616, but it was not a 
solution. I am now wondering if the only possibility is to clear the 
shared memory before restarting cflowd or there are more robust 
solutions..

Thanks 
Christian Cinetto
 



 



> 
> > Hi,  I have been running Cflowd for one month on two  routers Cisco 75XX.
> > I ran cflowdmux, cflowd , cfdcollect on the same dedicated machine, with
> > the following configuration:
> >
> > ...
> >
> > I tried to stop the processes and to start them back again; the behaviour
> > is more or less the same,i.e. after a while cfdcollect stops again.
> > I would appreciate any suggestion,
> 
> Have you checked:
> 
> http://www.caida.org/dynamic/archives/cflowd/0273.html
> http://www.caida.org/dynamic/archives/cflowd/0319.html
> 
> I solved this problem by commenting out the syslog call in 
> function CflowdCisco::ClearTableData() in file CflowdCisco.cc.
> 
> Regards
> Goncalo
> 

-- 
*******************************************
*	      Christian Cinetto           *
*    		                          *
*	   Network Operation Center       *
*           	  G A R R	          * 
* 	  christian.cinetto@garr.it       *
*                                         *
*              Fax 06/43361444            *
*	       tel.06/43361474            * 
*******************************************