We've had vaguely similar problems on Netview servers with ATM
interfaces. At the same time we were seeing OUT OF RESOURCES error when
running errpt | pg, and netstat -m was reporting mbuf overflows. Turns
out that there were two related problems- first that there is a memory
leak in the interface compression algorithm in base AIX 4.2.1. FIx for
this was to put latest AIX maintenance on for bos.net.tcp.client,
bos.net.tcp.server, bos.net.nfs.client, and bos.net.nfs.server. Second
problem was that the high potential throughput on the ATM interface
(which may be similar to the potential on the FDDI interface) required
an increase in the memory allocation to mbufs (set through the no
command variable thewall). mbufs buffer the interface traffic. This
thewall variable can be limited by the memory available on the machine.
To see the current thewall setting, run no -a | grep thewall. To
increase it, run no -o thewall=nnnnnn, where nnnnnn is the amount of
memory to allocate to mbufs (in kilobytes)
Remember that the no command only sets its variables until reboot, so to
make it permanent, you need to run this command in one of the bootup
config files. Usual location is rc.net, at the end of which the
tcp_sendspace and tcp_recvspace variables are set with the no command.
(no means "network options")
The mbuf tuning options on AIX v3 were much more extensive, but in AIX
4.x.x, the initial settings are system dependent, and the set through
thewall. It can be a little tricky. Read the man page for the no
command or run infoexplorer and search for thewall.
mbufs are automatically allocated from the pool across all interfaces,
so just having a fast interface can have an impact.
Hope this helps.
Gil Irizarry wrote:
>
> Leslie:
> Thank you for the suggestion. I 've change the default gateway back to the
> Ethernet interface because the system needed to be back in production. Now
> that it is back, I have a problem with netmon & nvpagerd. After applying &
> committing U453385. The netmon deamond is crashing every 10 minutes. I have
> opened a problem with AIX/NV6K support, however, I did not tell them about
> the nvpagerd problem.
> Is anyone out there having problems after PTF U453385 was applied?. I have
> accidentally committed the PTF, so, I would need to restore in order to undo
> the damage.
> NV support recommended increasing the buffer size to 2500, however it is at
> 25000 now. Is 25000 too much?
> HW 7025-F50, AIX 42.1 NV6K 4.1.2 (U453385) 512 meg memory.
> Thanks in advance
> Gil Irizarry
>
> >>> Leslie Clark <lclark@US.IBM.COM> 09/22/98 01:09pm >>>
> I have seen one case where pings at the commandline, or test... ping worked
> but
> netmon
> continually timed out, and when a sniffer was put on the segment with the
> management station
> we could see that the return pings were actually being lost, rather
> frequently.
> They were
> investigating the interface card and drivers; but in the meantime, it appeared
> that
> increasing the retries was more effective than increasing the timeout, since a
> dropped
> echo would never come back, but maybe the next one would. My gut feeling is
> that
> increased timeout is most helpful for snmp queries, and increased retries is
> most helpful
> for status polling. I don't claim to understand this, but that is how I treat
> it.
> Cordially,
>
> Leslie Clark
> IBM Global Services - Network & Systems Management - Detroit
>
> ----------------------------------------------------------------------------------------------------------------------------------------
>
> James:
> The timeout is set at 3.5 seconds for the routers. The routers are FDDI
> attached to the
> same Gigaswitch. I am going to shutdown the old interface & see what happens.
> Thanks you for the recommendation. I will let you all know the results.
> Gil
>
> >>> <James_Shanks@TIVOLI.COM> 09/18/98 12:26pm >>>
> You said pings and demand polls work fine, but if netmon is turning your
> interfaces red, it is because its pings are timing out. I suggest you run
> xnmsnmpconf and look at the default timeouts for these interfaces and see
> about raising the timeout a bit.
>
> James Shanks
> Tivoli (NetView for UNIX) L3 Support
>
> Gil Irizarry <IRIZARRG@LABS.WYETH.COM> on 09/18/98 10:56:32 AM
>
> Please respond to Discussion of IBM NetView and POLYCENTER Manager on
> NetView et alia <NV-L@UCSBVM.UCSB.EDU>
>
> To: NV-L@UCSBVM.UCSB.EDU
> cc: (bcc: James Shanks)
> Subject: Change Of NV6K IP ADDR
>
> Hi All:
> I have a RISC6K F50 with FDDI, Fast Ethernet & 10 meg Ethernet. I have
> changed the
> default gateway from the 10 meg to the FDDI nic. The FDDI connects to a
> gigaswitch
> concentrator that also houses the router with the IP address I am using as
> default.
> Now, NETMON is detecting "interface down" alerts frequently on all my
> routers
> interfaces. Shortly after, I see the "interface up" alert and all my router
> icon turn green
> again.
> I did run reset_ci after my change. Is there another utility needed?
> netstat -nr shows:
> "default xxx.xxx.xxx.xx UG 23 767159 fi0 -
> -" Where xxx is the IP
> address I have specified as the default gateway. Pings & demand poll work
> fine.
> The last entry on the display shows:
> "www.www.ww yyy.yyy.yyy.yyy.yyy U 159 771354 fi0 - -"
> Where www is the
> subnet and yyy is the IP address of my Netview box.
> Running:
> NV4.1.2 PTF LVL U453385
> AIX 4.2.1.4
> Thanks in advance
> Gil Irizarry
> 280 King of Prussia Road
> ST. Davids, PA 19087
> Phone: (610) 902-3426
> INTERNET: IRIZARRG@LABS.WYETH.COM
|