[Xymon] no client results arrive, lot of TIME_WAIT at server - SOLVED
Norbert Kriegenburg
norbert.kriegenburg at de.ibm.com
Fri Oct 20 13:39:22 CEST 2017
Hm,
this was a tricky one, but Thomas pointed me to the environment check which
then leads to the solution. Thanks!
I double checked the env for the xymon:
$ xymoncmd /usr/bin/env | egrep -i home
PATH=/home/xymon/client/bin:/bin:/usr/bin:/sbin:/usr/sbin:/etc
PWD=/home/xymon
HOME=/home/xymon
_=/home/xymon/server/bin/xymoncmd
XYMONHOME=/home/xymon/client
XYMONCLIENTHOME=/home/xymon/client
XYMON=/home/xymon/client/bin/xymon
XYMONTMP=/home/xymon/client/tmp
XYMONCLIENTLOGS=/home/xymon/client/logs
XYMONRUNDIR=/home/xymon/client/logs
LOCAL_CLIENTCONFIG=/home/xymon/client/etc/localclient.cfg
LOCAL_LOGFETCHCFG=/home/xymon/client/etc/logfetch.cfg
HOBBITCLIENTHOME=/home/xymon/client
BBHOME=/home/xymon/client
BB=/home/xymon/client/bin/xymon
BBTMP=/home/xymon/client/tmp
BBCLIENTLOGS=/home/xymon/client/logs
XYMONHOME points to the client dir, even if in xymonserver.cfg is defined:
XYMONHOME="/home/xymon/server" # The Xymon server
directory, where programs and configurations
So the serverprogs use the binaries from the client?
This also explains why some commands were not found or only worked with
relative path, because they do not exist in a usual client bin dir.
It was a fresh compiled xymon, so i checked the binaries.
I then found out that the XYMONHOME is compiled into the binaries. During
manual copy of the different progs into the client and server bin dirs (it
is a hardened environment were i don't have root access even for xymon
install) obviously i mixed up some client binaries into server bindir.
This then leads to those strange effects.
Now i replaced the server bins with the ones from the common dir in source
tree, and all is good.
$ xymoncmd /usr/bin/env | egrep -i home
PATH=/home/xymon/server/bin:/usr/local/bin:/usr/sbin:/usr/bin:/bin
PWD=/home/xymon/server/logs
HOME=/home/xymon
_=/home/xymon/server/bin/xymoncmd
OLDPWD=/home/xymon
XYMONHOME=/home/xymon/server
XYMONCLIENTHOME=/home/xymon/client
XYMONSERVERROOT=/home/xymon/server
XYMONSERVERLOGS=/home/xymon/server/logs
XYMONRUNDIR=/home/xymon/server/run
XYMONTMP=/home/xymon/server/tmp
HOSTSCFG=/home/xymon/server/etc/hosts.cfg
XYMON=/home/xymon/server/bin/xymon
I wonder if it is necessary to have those path hardcoded, or if this can be
replaced by generally using the environment vars?
Then we don't need to deal with different binaries for client and server
with the same name.
But as always, thx for the great tool and the help of the community!
Norbert
From: Norbert Kriegenburg/Germany/IBM
To: Thomas Eckert <thomas.eckert at it-eckert.de>
Cc: xymon <xymon at xymon.com>
Date: 19/10/2017 18:19
Subject: Re: [Xymon] no client results arrive, lot of TIME_WAIT at
server
Hi Thomas,
thank you for your hints.
Lets go through your suggestions:
- you were right with the INTERVAL. This was added by accident, but doesn't
harm (no more channels are started), removed, no effect
- yes, server process was restartet at 5:13am UTC
1.
- environment looks as expected:
$ xymoncmd /usr/bin/env | egrep -i 'xymsrv|xymonservers|bbdisp'
XYMONSERVERS=10.137.56.43
XYMSRV=10.137.56.43
BBDISP=10.137.56.43
BBDISPLAYS=10.137.56.43
2.
- unfortunately i don't have root access here for a tcpdump
3.
- now it gets strange:
if i try your suggestion i get this (doublechecked at other server and the
syntax works ok):
$ xymoncmd xymond_channel --channel=client cat
2017-10-19 15:58:58.843500 execvp() failed: No such file or directory
(but of curse xymoncmd is in the path)
if i try with relative path i get this (run from bin path):
$ ./xymoncmd ./xymond_channel --channel=client cat
2017-10-19 16:01:02.173601 Could not get shm of size 524288: No such file
or directory
2017-10-19 16:01:02.173943 Channel not available
some of these messages i can find in some logs:
$ grep "Could not get shm" *
alert.log:2017-10-19 10:38:00.733526 Could not get shm of size 1048576: No
such file or directory
history.log:2017-10-19 10:38:00.730686 Could not get shm of size 1048576:
No such file or directory
hostdata.log:2017-10-19 10:38:00.730538 Could not get shm of size 524288:
No such file or directory
rrd-data.log:2017-10-19 10:38:00.734198 Could not get shm of size 524288:
No such file or directory
I searched the web for those messages, and found some in hobbit/xymon
mailing list, but no details or solution.
http://lists.xymon.com/oldarchive/2008/06/msg00183.html
But: the client messages from server still arrive and the tests are all
visible...
very strange
Norbert
From: Thomas Eckert <thomas.eckert at it-eckert.de>
To: xymon <xymon at xymon.com>
Date: 19/10/2017 16:42
Subject: Re: [Xymon] no client results arrive, lot of TIME_WAIT at
server
Sent by: "Xymon" <xymon-bounces at xymon.com>
On 19 Oct 2017, at 14:49, Norbert Kriegenburg <
norbert.kriegenburg at de.ibm.com> wrote:
Yes,
that's what i checked at first. But i assume that also the info from
the servers own client data are captured by the same client channel
listener, and that obviously works.
valid point
Nothing in the clientdata.log, even if i activate the DEBUG mode with
a
kill -USR2 PID
My clientdata section:
# The client back-end module. You need this if you are running the
Xymon client on any system.
[clientdata]
ENVFILE $XYMONSERVERHOME/etc/xymonserver.cfg
NEEDS xymond
CMD /home/xymon/server/bin/xymond_channel --channel=client \
--log=$XYMONSERVERLOGS/clientdata.log xymond_client \
--no-clear-files \
--no-clear-ports \
--no-port-listing \
--uptime-status
INTERVAL 2m
The `INTERVAL` here is not im my and xymon.com’s-version of tasks.cfg— not
sure if it does harm (it _may_ try to start a new channel every 2 minutes).
ps listing:
$ ps -ef|grep "xymond_channel --channel=client"|grep -v grep
xymon 18865 18705 0 05:13 ?
00:00:00 /home/xymon/server/bin/xymond_channel --channel=client
--log=/home/xymon/server/logs/clientdata.log xymond_client
--no-clear-files --no-clear-ports --no-port-listing --uptime-status
If this process does not change it should be fine; your xymon server was
restarted 5:13 this morning?
I have a lot of TIME_WAITS from time to time, no idea why this goes
up and down:
$ netstat -atn|grep 1984
tcp 0 0 0.0.0.0:1984 0.0.0.0:* LISTEN
tcp 0 0 10.137.56.43:37752 10.137.56.43:1984 TIME_WAIT
tcp 0 0 10.137.56.43:37822 10.137.56.43:1984 TIME_WAIT
...
10.137.56.43 is the Xymon server IP.
$ netstat -atn|grep 1984|wc -l
165
(sometimes much much more)
(But goes down to 7 after some mins)
Looks like internal xymon-communication. As long as the TIME_WAIT-count
does not go >20k that should not be the cause.
Thing’s I’d try from here:
1. double-check the XYMSRV / XYMONSERVERS setting on the client(s), e.g.
with:
xymoncmd /usr/bin/env | egrep -i 'xymsrv|xymonservers|bbdisp’
2. just to double-check: tcpdump on xymon server and a client to check for
tcp/1984 communication
3. run a channel-listener an the server to see if the client messages
(other than the ones from the local client) make it
xymoncmd xymond_channel --channel=client cat | grep ‘^client '
xymoncmd xymond_channel --channel=client cat
The latter might be very noisy — the former just prints the “header”/1st
line of each connection.
There should be one message for each client sending data every 5 minutes.
HTH
Thomas
<graycol.gif>Thomas Eckert ---19/10/2017 11:14:56---Norbert, I assume
your manual (BB way) is something like `xymon 1.2.3.4 “status
example.com.test …”`
From: Thomas Eckert <thomas.eckert at it-eckert.de>
To: xymon <xymon at xymon.com>
Date: 19/10/2017 11:14
Subject: Re: [Xymon] no client results arrive, lot of TIME_WAIT at
server
Sent by: "Xymon" <xymon-bounces at xymon.com>
Norbert,
I assume your manual (BB way) is something like `xymon 1.2.3.4
“status example.com.test …”`.
The xymon-client send `clientdata` that is processed by a “channel
listener” and _that_ sends status-messages similar to above (a bit
simplified).
That channel-listener may not be running/crashing:
Check our `tasks.cfg` for the section `[clientdata]` and verify, if
the `xymond_channel`-process is running (I suspect it’s not). Maybe
the `clientdata.log` helps with debugging. You may also run the task
manually to get further information.
What’s your problem with the "lot of TIME_WAIT”? Are these just for
tcp/1984-connections?
Cheers
Thomas
On 19 Oct 2017, at 07:30, Norbert Kriegenburg <
norbert.kriegenburg at de.ibm.com> wrote:
experts,
i have a new (not my first) Xymon installation (4.3.28 on
RH 7.4, self compiled).
All looks good, but the clients only show the network
tests (conn, ssh). Only the Xymon server itself shows
everything (disk, cpu,...).
- all clients are RH 7.4, same install image used (i just
copied over the client part with client config, create a
cronjob for automatic start, so no root access nor extra
install needed)
- nothing suspicious in server or client logs (even in
debug mode)
- no ghosts
- if i send a test manually (the old BB way) this shows
at Xymon immediately (so the client settings are ok)
- if i shutdown the xymond at server, the client logs
complains:
2017-10-19 05:09:18.051994 Whoops ! Failed to send
message (Connection failed)
2017-10-19 05:09:18.052104 -> Could not connect to Xymon
daemon at 10.137.56.43:1984 (Connection refused)
(so obviously they communicate)
Querying the xymond shows:
xymon 127.0.0.1 "query clientname.ssh"
green <!-- [flags:OrdastLe] --> Thu Oct 19 05:21:29 2017
ssh ok
is ok, but
xymon 127.0.0.1 "query clientname.disk"
shows nothing
Any ideas?
Norbert
_______________________________________________
Xymon mailing list
Xymon at xymon.com
http://lists.xymon.com/mailman/listinfo/xymon
_______________________________________________
Xymon mailing list
Xymon at xymon.com
http://lists.xymon.com/mailman/listinfo/xymon
_______________________________________________
Xymon mailing list
Xymon at xymon.com
http://lists.xymon.com/mailman/listinfo/xymon
_______________________________________________
Xymon mailing list
Xymon at xymon.com
http://lists.xymon.com/mailman/listinfo/xymon
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xymon.com/pipermail/xymon/attachments/20171020/35a58428/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: graycol.gif
Type: image/gif
Size: 105 bytes
Desc: not available
URL: <http://lists.xymon.com/pipermail/xymon/attachments/20171020/35a58428/attachment.gif>
More information about the Xymon
mailing list