Conference smurf::ase

1988.0. "No buffer space available!!!!." by MDR01::CARRANZ (MCS Madrid) Mon Apr 07 1997 10:02

Hello,

	

	We have a DECsafe environment v1.3 with 2 network interfaces:

		FDDI
		Ethernet interface.

	Primary network for ASE is the FDDI, Ethernet has been configured
as backup.

		2 * 4100 with DU 3.2F.

	Our problem was that the other day, nobody could access 
to ASE service defined on one system.

	The daemon.log in that moment gave next messages:

Apr  1 10:08:18 rtrprd2 DECsafe: local HSM Warning: Can't ping rtrprd1-f over th
e network
Apr  1 10:08:20 rtrprd2 DECsafe: local HSM Notice: /var/ase/sbin/ase_run_sh: cha
nge host 10.10.47.234: gateway 10.10.33.220
Apr  1 10:08:20 rtrprd2 DECsafe: local HSM Notice: /var/ase/sbin/ase_run_sh: cha
nge host 10.10.33.220: gateway 10.10.33.220
Apr  1 10:08:20 rtrprd2 DECsafe: local HSM ***ALERT: HSM_PATH_STATUS:10.10.33.22
0:UP:10.10.47.234:DOWN
Apr  1 10:54:48 rtrprd2 DECsafe: rtrprd2 Agent Notice: stopping service rtrprd
Apr  1 10:56:27 rtrprd2 DECsafe: rtrprd2 Agent Notice: /var/ase/sbin/ase_run_sh
rtrprd: rtrprd: stop at Tue Apr  1 10:54:49 MET DST 1997
Apr  1 10:56:28 rtrprd2 DECsafe: rtrprd2 Agent Notice: /var/ase/sbin/ase_run_sh
rtrprd: RTRPRD: Parando la aplicacion
Apr  1 10:56:28 rtrprd2 DECsafe: rtrprd2 Agent Notice: /var/ase/sbin/ase_run_sh
rtrprd: stty: tcgetattr: Not a typewriter
Apr  1 10:56:28 rtrprd2 DECsafe: rtrprd2 Agent Notice: /var/ase/sbin/ase_run_sh
rtrprd: ping: sendto: No buffer space available
Apr  1 10:56:28 rtrprd2 DECsafe: rtrprd2 Agent Notice: /var/ase/sbin/ase_run_sh
rtrprd: PING rtrp-f (10.10.47.244): 56 data bytes
Apr  1 10:56:28 rtrprd2 DECsafe: rtrprd2 Agent Notice: /var/ase/sbin/ase_run_sh
rtrprd: ping: wrote rtrp-f 64 chars, ret=-1 


	At that moment, you could do a ping at FDDI computer address, but you
couldn't access to any FDDI ASE service.

	When you try to do a ping to an ASE FDDI service it said:
    
    	"No buffer space available".
    
    	Ethernet works OK.
    
    	We didn't see any messages saying that "fta0: Link unavailable".
    
    	We did:

	# ifconfig fta0 down
	# ifconfig fta0 up

	Doing that, the problem was solved, but this is a temporary
    solution, and i would know which is the real problem customer has.
    
    
	In order to solve the problem i don't know if this could be a system
parameter or perhaps a configuration problem from dec safe. If anybody knows
which parameter, please let me know.

	I'm going to analyze with "netstat -m" and "vmstat -M" the status
when there is no problem.

	I'm going to suggest customer to do a procedure that do a ping
to other computer thru Ethernet/FDDI and each other thru both interfaces and
to ASE services address, in order to isolate the problem.

	Any other suggestion would be appreciatte.

	
	Thanks in advance,


	Carmen Arranz.

	MCS. Madrid. Spain.

	
	**************  Cross posted in ASE/DIGITAL UNIX  **************


    

    Please, has anybody out here any idea about this problem?
    
    Any help will be very appreciate.
    
    Thanks in advance,
    
    
    Carmen Arranz
    

Hello Carmen,

I don't know which interface is your FDDI, but according to this one of the
interfaces (10.10.47.234) was seen as DOWN:

Apr  1 10:08:20 rtrprd2 DECsafe: local HSM ***ALERT: HSM_PATH_STATUS:10.10.33.22
0:UP:10.10.47.234:DOWN

If the FDDI goes down, it is possible to see the error message "no buffer space
available".  This is because even an unconnected interface will queue up 
packets.  Eventually the send queue for the interface will fill up, and you
will get this error.

There is a patch coming that will help in this case.  If you need it sooner, 
you can open a CLD and I will get you a test patch.  But, what you really
need to figure out is why you lost your FDDI connection.  Fix the problem 
with the FDDI, and you won't have the "no buffer space..." problem.

Rob Marshall
USEG

	Thanks for your reply,

	Sorry, 10.10.47.234 is the FDDI network address, as you have guessed.

	We have had  some ASE messages about that FDDI is down, and after some 
seconds it's up.

	Today we have had a big customer problem, i suppossed related with 
this one:

	Customer couldn't connect to the ase alias services, it wasn't 
responding, but you could do a ping to that physical interface "10.10.47.234":

	# ping 10.10.47.234 was responding

	And customer could do ping to any FDDI network node without any problem.

	May be that possible?

	Few seconds before that, customer had the same message about that ASE 
couldn't do a ping thru FDDI, FDDI was down (same as .0).

	I have read that when you have primary network saturated, and if you
out of mbufs, you should adjust the "ubcmaxpercent" and "ubcminpercent".

	My question now is that i'm not sure if here we are having one problem
or two different problems.

	Could anybody help me in that question?.

	Now customer will safe into a file then "netstst -in", "netstat -m"
and "vmstat -M" outputs daily and will send them to me.

	We have raised every ASE processes and we have increased too the 
timing parameters into "/etc/hsm.conf".

	We have increased next parameters:

	HSM_PNET_TBS=100
	HSM_PNET_TBF=120
	HSM_PNET_MR=4

	rest parameters have default values.

	How could i delimited the FDDI ASE problem.

Any help would be very appreciate.
    
    Thanks in advance,
    
    Carmen Arranz