LNETSTAT(1,C) AIX Commands Reference LNETSTAT(1,C)
-------------------------------------------------------------------------------
lnetstat
PURPOSE
Provides network statistics.
SYNTAX
+-----------------+ +------------- -p ------------------+
lnetstat ---| |---| +------+ +---------------------+ |---|
+- -s --- site ---+ +-+- -S -+--| +------------+ |-+
^ | +- -R -+ +- -t -| |-+
+--------+ +- interval -+
DESCRIPTION
The lnetstat command provides statistics about TCF message traffic. If no
arguments are specified, lnetstat displays network information about all sites
in the current cluster except the local site. The local site network message
statistics are ignored if the site number is specified or when all site
information is printed. This information includes site number, name, and how
long the site has been up.
If TCP/IP message traffic is required, use the netstat command.
The site parameter values define the site to which network messages are routed,
the timeout value for transmissions, the number of retries (re-transmissions)
for a network message, the acknowledgement windows for normal and special
messages, and whether check summing is on (1) or off (0).
Site status information describes the status of network messages coming from
and going to a specific site, whether or not the site is accepting messages,
messages that have been acknowledged or dropped, re-transmitted messages, and
time in contact with foreign sites.
Message statistics for each site include statistics for the number of incoming
and outgoing messages and their response time for various classes of network
messages handled by the site.
FLAGS
-p Prints the site parameters; these are values used by the kernel
in handling all traffic to and from a given site.
-R Displays re-transmission counts to all sites (default) or to
those specified with the -s flag.
Processed November 8, 1990 LNETSTAT(1,C) 1
LNETSTAT(1,C) AIX Commands Reference LNETSTAT(1,C)
-s site Restricts display to the specified sites. Sites can be
specified either by name or number, and should be separated with
a space. If the local site is specified, it is silently
ignored.
-S Summarizes; message counts and times are summed over all sites
(default) or those specified with the -s flag.
-t interval Displays new traffic once every interval seconds. If interval
is not specified, it defaults to 10.
EXAMPLES
1. The command
lnetstat -p -s myoko
displays the following statistics:
site 5 myoko
route: 5 timeout: 4 retries: 3
window: 6 s_window: 2 checksum: 1
2. The command
lnetstat -s myoko
displays the following statistics:
site 5 myoko connected for 0d 01h 03m 58s
SITE STAT MSG_STAT msgin msgout rsptime
=====================================================================
chan_stat open misc 164123 127103 12.5
site_stat up nops 85027 89789 0.2
in_pkts 763636 read 77238 81979 10.5
out_pkts 761604 writes 51 10117 19.1
dups 352 open, stat 78942 122325 16.0
retrans 585 rcd ops 448 132 8.3
msgdrops 2 synch 0 0 -
bufdrops 0 proc crt/dst 13 11 116.3
xmterrs 0 proc sig/sts 18 8 10.0
crcerrs 0 cls/cmt/trunc 78985 81124 923.8
badmsg 0 read rsp 94410 77517 9.7
badsite 2 other rsp 174095 168027 7.7
badseq 103 css-ss 4 24 5.0
ack_blocks 382 topchg/update 18 25 25548.8
sack_blocks 3 netwrk maint 9934 10062 6.2
chan_opens 28 user-1 0 0 -
channel 409 user-2 0 0 -
user-3 0 0 -
user-4 0 0 -
untimed msgs 18 user-5 0 0 -
Processed November 8, 1990 LNETSTAT(1,C) 2
LNETSTAT(1,C) AIX Commands Reference LNETSTAT(1,C)
3. The command
lnetstat -S -s myoko
displays the following statistics:
Summary of site(s) 5
SITE STAT MSG_STAT msgin msgout rsptime
=====================================================================
chan_stat xxxxxxx misc 164126 127106 12.5
site_stat xxxxxxx nops 85028 89790 0.2
in_pkts 763646 read 77239 81980 10.5
out_pkts 761614 writes 51 10117 19.1
dups 352 open, stat 78943 122326 16.0
retrans 585 rcd ops 448 132 8.3
msgdrops 2 synch 0 0 -
bufdrops 0 proc crt/dst 13 11 116.3
xmterrs 0 proc sig/sts 18 8 10.0
crcerrs 0 cls/cmt/trunc 78986 81125 923.8
badmsg 0 read rsp 94411 77518 9.7
badsite 2 other rsp 174097 168029 7.7
badseq 103 css-ss 4 24 5.0
ack_blocks 382 topchg/update 18 25 25548.8
sack_blocks 3 netwrk maint 9934 10062 6.2
chan_opens 28 user-1 0 0 -
channel xxxxxxx user-2 0 0 -
user-3 0 0 -
user-4 0 0 -
untimed msgs 0 user-5 0 0 -
4. The following is an example of inetstat information for a site called
sonic:
Processed November 8, 1990 LNETSTAT(1,C) 3
LNETSTAT(1,C) AIX Commands Reference LNETSTAT(1,C)
site 2 sonic connected for 13d 10h 49m 46s
SITE STAT MSG_STAT msgin msgout rsptime
=====================================================================
chan_stat open misc 3735 1807 21
site_stat up nops 13371 44918 1
in_pkts 115059 read 309 1873 84
out_pkts 144272 writes 35 1152 24
dups 0 open, stat 724 6762 47
retrans 8 rcd ops 2775 32764 16
msgdrops 0 synch 0 0 0
bufdrops 0 proc crt/dst 421 462 26
xmterrs 0 proc sig/sts 308 710 117
crcerrs 0 cls/cmt/trunc 429 3106 157
badmsg 0 read rsp 4510 309 85
badsite 0 other rsp 77672 10655 23
badseg 0 css-ss 0 606 13
ack_blocks 0 topchg/update 212 219 24
sack_blocks 0 netwrk maint 9580 38731 13
chan_opens 1
channel 194
untimed msgs 4
The SITE STAT column describes the site status along with associated
information in the next column. The MSG_STAT column describes the message
status type. msgin gives the number of messages in, msgout gives the
number of messages out, rsptime gives the response time.
Each site has a chart of SITE STAT and MSG_STAT information, which
indicates the following:
SITE STAT
chan_stat Channel status (open, close, attempt).
site_stat Status of other site (up or down).
in_pkts Number of TCF messages received.
out_pkts Number of TCF messages sent.
dups Number of TCF messages received (three are counted as two,
one worked).
retrans Number of TCF messages sent (three are counted as two, one
worked).
msgdrops When TCF message was received there were no free netmsgs
available, so the TCF msg was dropped.
bufdrops When long message was received, there were no free
filesystem buffers, so TCF msg was dropped.
xmterrs Transmit errors; no TCF msg may have been transmitted; error
in lower-level protocol.
crcerrs Checksum error in TCF header.
badmsg Malformed message (too short, consistency check bad).
Processed November 8, 1990 LNETSTAT(1,C) 4
LNETSTAT(1,C) AIX Commands Reference LNETSTAT(1,C)
badsite TCF site number in TCF msg was not the same as the TCF site
that received it for previous n messages had yet been
received*.
badseq Bad sequence, not in order.
ack_blocks Number of times message ready to be sent but could not be
sent, as no acknowledgement.
sack_blocks Number of times special message (topchange) ready, not sent.
chan_opens Number of times channel established or attempted.
channel Channel number (incremented after it goes down).
MSG_STAT
misc Messages not included in any other category.
nops No operations, as it carried no information (only an
acknowledgement).
read Number of filesystem reads.
write Number of filesystem writes.
open, stat Number of open and stat calls.
rcd, ops Remote character device terminal operations reads, writes,
I/O controls.
proc crt/dat Process creation and destruction messages.
proc sig/sts Process signal and status message.
cls/cmt/truc Filesystem (not remote terminal) close, commit, or
truncated.
read rsp Read response (on filesystem, not on remote terminal).
other rsp Responses other than read for filesystem messages.
css-ss Controlling storage site and storage site message.
topchg/update Topology change related.
netwrk maint Channel opens, probes, responds to IP level of TCF protocol.
* where n is the window size
Note the distinction between filesystem operations and remote terminal
operations. The key to understanding how to use the Inetstat parameters is
to watch for patterns in network traffic every day and then be aware of the
correlation between the various parameters and the changes that the cluster
is undergoing. It is also useful to check the TCP/IP traffic (using
netstat) and correlate TCP/IP changes with what is going on with TCF
changes. For example, if drops are higher (say 500 in one hour, where
normally you have 100 per day or so), this indicates insufficient resources
and therefore more netmsgs need to be configured in /etc/system. (If
netparams are not listed, then check the default /etc/master file for this
number.)
If dups and drops go up simultaneously, then one side is losing messages.
If dups increase, but drops do not the site is too slow and is losing
messages.
If retrans is high, the site is sending lots of messages out. This could
be caused by no buffers, hardware being wedged, or a cable problem.
Processed November 8, 1990 LNETSTAT(1,C) 5
LNETSTAT(1,C) AIX Commands Reference LNETSTAT(1,C)
If dups and retrans constitute a large percentage of inpkts and out_pkts,
then further analysis is warranted to find out why transmissions are not
working.
High crcerrs is due to a checksum problem, which is usually a hardware
problem.
High badmsg counts are often due to a hardware or kernel problem. Look at
other indicators for more information.
badseq, dups, and retrans are all related, since bad messages give
out-of-sequence errors and numbers.
FILES
/usr/include/sys/netstat.h
/usr/include/sys/netctrl.h
RELATED INFORMATION
In this book, see the following command: "crash."
See the AIX/370 Diagnosis Guide.
Processed November 8, 1990 LNETSTAT(1,C) 6